Social networks in special schools: Friendships, bullying and play among pupils with special educational needs and disabilities

Received: 24 January 2026

® Check for updates

Accepted: 12 August 2026

Dol: 10.1111/1467-8578.70121

Original Article

Nasen

Social networks in special schools: Friendships, bullying and play among pupils with special educational needs and disabilities

Julia R. Badger'® | Lucy Robinson'® | Kimberly Winardi'® | Ariel Lindorff'® |

Chloe Bracegirdle”* ©

'Department of Education, University of Oxford, Oxford, UK 2Department of Sociology, University of Oxford, Oxford, UK

*School of Psychology, University of Surrey, Guildford, UK

Correspondence Julia R. Badger, Department of Education, University of Oxford, Oxford, OX2 6PY, UK.

Email: julia.badger@education.ox.ac.uk

Funding information John Fell Fund, University of Oxford, Grant/Award Number: 0015442

Abstract

Little is known about the social networks of children with special educational needs and disabilities (SEND) in specialist school settings—schools designed to support specific educational needs. Most literature on the social connections of children with SEND suggests that they have fewer friendships and are more involved in bullying situations, but these data are usually collected from mainstream school settings, where children with SEND are the minority. To better understand the social connections of pupils with SEND in specialist school settings, we conducted a social network study in this educational context. Specifically, this research studied whole-school social networks in three special schools with differing specialisms: (1) social, emotional and mental health; (2) cognition and learning; and (3) autism. Positive peer networks (friends and play companions) and negative peer networks (bullying and victimisation) largely mirrored the social networks of mainstream schools, although there was less student hierarchy. However, we also identified important differences between the special school types, such as fewer connections overall in the social, emotional and mental health school. We offer novel insights for practitioners into the social networks of children with SEND. Our methodology confirms the feasibility of social network analysis in the special school context, opening avenues for future longitudinal research.

Keywords bullying, friendship, play, schools, social network, special educational needs and disabilities

Key points

* It is possible to successfully conduct social network studies with special school pupils.

* Pupils in special schools show similar numbers of friendships as those without SEND in mainstream schools.

* Pupils in special schools show similar negative social connections to those with- out SEND in mainstream schools.

+ Fewer positive and negative peer connections were found in the social, emotional and mental health school compared to the autism and cognition and learning schools.

This is a access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,

provided the original work is properly cited.

Br J Spec Educ. 2026;00:1-14.

wileyonlinelibrary.com/journal/bjsp_ | 1

Badger er Al.

*|nasen

Introduction

Children and young people spend a large proportion of daily life with their peers in school. The social con- nections formed in school play important roles when it comes to feelings of identity, belonging, social status and mental health. Positive connections, such as friend- ships, can support resilience, healthy psychosocial func- tioning and higher academic achievement (Bagwell & Bukowski, 2018; Van Harmelen et al., 2017), whereas negative connections, such as bullying, can lead to feel- ings of isolation, lowered self-worth and depression (Christina et al., 2021; Schoeler et al., 2018).

Social networks in schools

Most commonly, studies considering peer relationships in schools have done so from a singular perspective, fo- cusing on the individual without considering the wider social network within the school environment. This can take the form, for example, of an individual self- reporting how many friends they have, or how often in the past few months they have been bullied. One alter- native method of studying the structure, distribution and formation of peer relationships, though, is through social network analysis (Sna), which uses peer nomi- nations to map the connections among all individuals in a given social context (e.g. a school). This approach moves away from the simplification of an individual focus to allow the identification of connections between multiple pupils, which is important for understanding peer group dynamics such as friendships and bully- ing. This approach can provide additional depth to the information.

Studies applying the social network approach have revealed much about peer relationships in mainstream schools. In a study of 848 classroom networks—17,469 pupils aged 14-15—from five European countries, the average number of friendships a person said they had (friendship nominations) was 3.5 with an average re- ciprocation rate (when a nominated friend returned the nomination of friendship) of 51% (break down for England: 3.1 and 45%; Vo6rds et al., 2019). However, al- though positive peer connections, such as friendships, are often reciprocated, pupils are less likely to share perspectives regarding negative connections, such as bullying involvement. Tatum and Grund (2020) analysed social network data from 96 school classes in Germany and found that only 26% of the bullying accusations made by participants (accusing another child of bully- ing them) were confirmed by a matched bullying con- fession. They also found that only 18% of confessions (of bullying someone else) were confirmed with a matched accusation, showing substantial reporting discrepancies between bullies and victims.

Of course, beyond nomination reciprocation, are the pupils who neither nominate nor are nominated by oth- ers: the isolates (individuals who have zero peer connec- tions). In a recent study of 981 UK pupils aged 11-14, 99% of pupils reported having one or more friends and/ or were nominated as a friend by one or more of their peers, leaving only 1% of pupils within the isolate cate- gory (Bracegirdle et al., 2023). Thus, a small percentage of individuals have zero friendship connections, while most individuals will have only a few friendship con- nections, and only a few individuals will have a lot of friendship connections (Gonzalez et al., 2007; Rivera et al., 2010). Having more friendship connections can be a potential indicator of influence or social status among peers. Interestingly, Sna has also shown bul- lies to have higher centrality (Salmivalli et al., 1997) and in support of this, Pakaslathi and Keltikangas- Jarvinrn (2001) found that both positive and negative behaviours have been found to be linked to a pupil's level of centrality (popularity) in a network. When it came to negative connection reporting from the same 981 UK pupils, only 36% of pupils gave or received a bullying involvement nomination, leaving 64% of pupils as isolates (Bracegirdle et al., 2023). This number mir- rors research from 40 countries that states an average of 37%—42% of pupils report bullying (World Health Organisation, 2020). Bullying connections are much rarer than friendships connections.

Social network studies conducted in mainstream schools have also shown high levels of transitivity. Transitivity is the tendency for people to become friends with the friends of their friends, resulting in closed three- way connections (Rivera et al., 2010). This may occur for various reasons. For example, individuals with a mutual acquaintance are likely to meet and spend time together (Granovetter, 1973), and individuals are inclined to feel positively towards their friends' friends, seeking bal- ance in their relationships (Heider, 1946). Among many social contexts, transitivity has been observed in main- stream school friendship networks, including a study of 11 schools in the UK and 23 schools in the Netherlands (Block, 2015).

Studies have also shown that social networks in main- stream schools are often segregated across various so- ciodemographic dimensions (e.g. gender, ethnicity), as pupils tend to form social connections with those who are similar to them (McPherson et al., 2001; Moody, 2001). Segregation has been observed in both negative and positive connections (e.g. 86% same-ethnic friendship connections, 82% same-ethnic bullying connections; Bracegirdle et al., 2023). Overall, research has revealed much about the social networks of pupils in mainstream schools. However, less is known about the social net- works of children with special educational needs and disabilities (SEND), particularly those in specialist edu- cation schools (hereafter special schools).

fuaseujp'sdny t03y papeorun

joney auesy90;9 ystueds kq {Ziol'SLSS-LOFI/Ttt Ot topauos:Koqr-Arexqyyautt

ypu pu suLAY, ap 295 [9ZOZIOT/To] Ho AreIq_T Auntue K>qEAK “(Pepiurs ap OUArSTUTTA) WoIS

jqvandde xp £q pou9908 are sajanime yo :95n Jo saqna 10p Kresge] Auntie AoqEAK Wo (stonsPDD-pur-suATaYUIOD Capea Aeg

astta0r7 stot ane:

Social Networks In Special Schools

nasen|?

Special educational needs and disabilities (SEND)

As of June 2025, over 1.7 million pupils in England (ap- proximately 18%) had a noted SEND status, a 5.6% increase from 2024 (Gov.UK, 2025). Of these, approxi- mately 160,000 pupils attended a special school.

The UK sets out four broad areas of need within the SEND code of practice (Department for Education, 2026): (1) communication and interaction, which includes autism, (2) cognition and learning, (3) social, emotional and mental health, (4) sensory or phys- ical needs. These four broad areas of need encompass seven specific types of specialist schools (although many schools cater to more than one specialism): (1) speech, language and communication, (2) autism, (3) cognition and learning (significant), (4) cognition and learning (moderate), (5) social, emotional and mental health, (6) sensory, (7) physical. Each school will provide specialist knowledge, support and teaching in line with the pri- mary needs of their pupils.

Friendships and bullying of children with SEND

For many children with SEND, attending school is the main opportunity for them to connect with many peers their own age and trial the formation of social connections by choice (as opposed to by convenience with siblings or family friends). Yet these children can find forming and maintaining friendships more challenging, a scenario which can lead to heightened vulnerability (Campbell et al., 2019; Zablotsky et al., 2013). Research often shows that children with SEND have fewer friends than their non-SEND peers (Heiman, 2000; Pijl et al., 2008), and that a higher percentage will be involved in bullying (up to 69%; Rose et al., 2011). A study by Chamberlain et al. (2007) mapped the social networks of 17 autistic children in mainstream classrooms and although they found that these pupils did not report greater levels of loneliness or solitude among their peers, their network data revealed that they had far fewer friendship con- nections and lower centrality and reciprocation than that of their non-autistic peers. Highly similar findings from mainstream schools were later reported by Locke et al. (2010) in their sample of seven autistic pupils and Pinto et al. (2019) in their sample of 59 pupils with SEND, both also showing poorer friendship quality, lower so- cial status, fewer reciprocated friendships and less peer integration.

Most studies reporting the connections of children with SEND focus on a small sample of pupils in main- stream settings. Yet, it is possible that social networks may differ between mainstream and special schools given the different student populations and school character- istics. For example, compared to mainstream schools, special schools typically have fewer pupils, a higher

staff-to-pupil ratio, more flexible attendance hours, pu- pils with more complex needs, and more time allocated to breaks, play, and interactive tasks, all of which likely influence opportunities for social connection formation and maintenance. A non-SNA qualitative study by Holt et al. (2017) considered the social encounters of 46 pupils aged 11-16 with SEND within two mainstream schools and two special schools. They found that bullying oc- curred in all schools but that although pupils with SEND in all settings had friendships, these were more prevalent within the special school setting than the mainstream school setting.

A difficulty when trying to understand the social networks of children with SEND, is that they are often grouped as one: ‘SEND’, where in reality their social understanding, communication capabilities and rec- ognition of others! intention can vary greatly. Another difficulty when trying to understand the social networks of children with SEND is that many of these children will not be able to access, understand or respond ac- curately to the traditional methods of data collection (typically independent self-reporting via word-dense questionnaires).

Current study

This research presents the first full social network study conducted in the special school context. We collected whole-school peer network data from three English special schools, each with a differing special- ism—(1) social, emotional and mental health (Semh), (2) cognition & learning (C&L), (3) autism—to map friendship, play and bullying connections within the schools as well as to look for similarities and differ- ences between the schools. We also explored the fea- sibility of gathering meaningful social network data from pupils in those settings.

Our study was therefore guided by the following questions:

1. Is it possible to collect social network data from pupils in special schools, given the nature of pupils' potentially complex needs?

2. What do positive social networks (friendship and play) look like in special schools?

3. What do negative social networks (victimisation and perpetration) look like in special schools?

Method Design A cross-sectional design was used, working with as many

children within each school as possible between January and March 2025.

fuaseujp'sdny t03y papeorun

joney auesy90;9 ystueds kq {Ziol'SLSS-LOFI/Ttt Ot topauos:Koqr-Arexqyyautt

ypu pu suLAY, ap 295 [9ZOZIOT/To] Ho AreIq_T Auntue K>qEAK “(Pepiurs ap OUArSTUTTA) WoIS

jqvandde xp £q pou9908 are sajanime yo :95n Jo saqna 10p Kresge] Auntie AoqEAK Wo (stonsPDD-pur-suATaYUIOD Capea Aeg

astta0r7 stot ane:

Badger er Al.

‘lnasen

Participants

We collected data from three English special schools, a total of 156 pupils.

School I's specialism was Semh with an age range of 11-16 (M= 13.6; Sd=1.41); 5% female. Sixteen pupils did not assent to participate and seven were absent throughout the data collection period. This resulted in a final partici- pation number of 39/62 pupils enrolled at the school (63%).

School 2's specialism was C&L with an age range of 10-16 (M=12.5; Sd=1.60); 41% female. One pupil was opted out by their parents, five pupils did not assent, two were too dysregulated to participate, and one was ab- sent. This resulted in a final total of 58/67 pupils (87%).

School 3's specialism was autism with an age range of 11-16 (M=13.2; Sd=1.43); 39% female. Two pupils did not assent and one was absent. This resulted in a final total of 59/62 pupils (95%).

Social network measures

Four questions were asked of each participating pupil: “Think about all the children in your school,

Who are you friends with? . Who do you play with?

Who is mean to you?

. Who are you meant to?”

Reve

In response pupils could nominate up to 10 peers by name. The term ‘mean’ instead of ‘bullying’ was used for two reasons. Firstly, several special schools do not use the term bullying and so, even with a provided defini- tion, the children may not be as clear on the meaning of this terminology. Secondly, the word mean is often used to infer bullying, and distinguish from other behaviours (Volk et al., 2014).

Procedure

All research was carried out in line with the British Educational Research Association Ethical Guidelines (2024). Schools opted into the study; parents/guardians were first informed about the study and could then opt their child out of the study if they preferred them not to participate. All pupils were asked whether they assent to participate on a given day. For those not opted out, the schools provided pupil names, age, sex and class group. Participants and their parents/guardians were made aware that they could withdraw from the study and have any data already collected removed and destroyed, up until the point of full anonymisation (specific dates were given) without giving a reason. Researchers were also alert to any non-verbal signs that participants no longer wished to participate.

Schools were sent social stories to show to their pupils before the researchers' visit. These presentation slides explained who the researchers were (including photo- graphs), why they would be visiting, what they would be asking of the pupils, and what we would be using the data for. The researchers spent a few hours in each school before data collection so pupils could become fa- miliar with them.

Data collection took place over the equivalent of eight school days. Each session took approximately 10min and each pupil worked individually with a pair of researchers in an interview-style format to provide answers to the four questions. One researcher would read out the questions while the other would record the pupil's verbal peer nominations. In most cases, the pupil and researchers were in a separate, quiet space close to the classroom. In a few cases, pupils only felt able to participate when they remained within the classroom. During the interview, data were entered using an electronic tablet into an excel spreadsheet pre-populated with details of every participating pupil in the school to ensure accurate reporting. In the case where a non-participating pupil name was given, the researcher recorded ‘unknown’. In the case where a pupil outside of the school was named, no data was recorded. Once complete, all data were anonymised.

Results Network visualisations

The following figures show the network visualisations for each school (see Figure | for Semh; Figure 2 for C&L and Figure 3 for autism). Each figure includes four pan- els, showing the (a) friendship network, (b) play network, (c) victimisation network, and (d) perpetration network. Within each panel, a node represents a pupil (square=a girl) and each arrow linking nodes represents a nomina- tion: the arrow points from the nominator to the nomi- nee. If the arrow is double-headed then the nomination is reciprocated. The palest shade represents the youngest pupils (aged 10); the darkest shade represents the oldest pupils (aged 16).

Descriptive network statistics

Table 1 presents descriptive network statistics for each school.

Positive network characteristics:

Friendship and play

Across schools, pupils nominated around two to four peers (mean outdegree: 1.87—4.27), 41%-64% of

fuaseujp'sdny t03y papeorun

joney auesy90;9 ystueds kq {Ziol'SLSS-LOFI/Ttt Ot topauos:Koqr-Arexqyyautt

ypu pu suLAY, ap 295 [9ZOZIOT/To] Ho AreIq_T Auntue K>qEAK “(Pepiurs ap OUArSTUTTA) WoIS

jqvandde xp £q pou9908 are sajanime yo :95n Jo saqna 10p Kresge] Auntie AoqEAK Wo (stonsPDD-pur-suATaYUIOD Capea Aeg

astta0r7 stot ane:

Social Networks In Special Schools

(a) Who are you friends with?

e ° @ e _ * | e e & @ ° ° © ° e ° °. ° e ~~» @ e > * ee. ?

(d) Who are you mean to?

Figure 1

nominations were reciprocated, 21%-51% of triplets were closed (transitivity), and only 0%-14% of pupils were isolates. Positive connections were relatively evenly distributed across pupils (centralisation) with a proportional range of 0.12—0.20, suggesting no clear hierarchy between pupils in any of the schools.

The autism school recorded the most friendship connections (M@=4.27, Sd=2.19) and play connec- tions overall (M=3.02, Sd=2.49), with about half of these reciprocated (0.51 and 0.42, respectively). The C&L school had slightly fewer friendship connections (M=3.86, Sd=2.34) and play connections (M=2.60, Sd=2.14) but showed the highest level of reciproca- tion (0.60 and 0.64, respectively). The Semh school

Network visualisations for the Semh school. The girls (1=3) have been removed due to potential identifiability.

showed the fewest friendship connections (M=2.69, Sd=1.96) and play connections (M=1.87, Sd=1.59) as well as the lowest level of reciprocation (0.48 and 0.41, respectively).

The C&L school showed a slightly higher proportion of transitivity in both friendships and playmates com- pared to the autism school (C&L: 0.44 and 0.51, respec- tively; autism: 0.39 and 0.48, respectively) suggesting more tight-knit friendship groups, and the Semh school showed lower levels of transitivity especially for play- mates (0.33 and 0.21, respectively).

The Semh school was the only school to record friendship isolates: two children (5%) in this school neither named a friend nor were named as a friend.

vojesdnw) SuorIpUED pur Suiay, AP 295 “[9ZOTIOL/Io] Wo AmEIqr'T Suttud Kaila “(Pepites ap Oudistuipa) Hoistaeia Tpuonen, aueIyDOD Ysis Kq Iztol’BLSS-LOFV/Ltlhol/opAWOD Kayo KreaqH Auto" seusMofuaseUjp:sdny tI} PapEOTE

sop Armage'y auHTUG) KaILAA Ho (StOR|PUED-puw-suAIOY/Ulo> K9A-ATEIGH

rqvandde axp kq pous9908 are sajone Yo :2sn 5

astta0r7 stot ane:

Badger er Al.

a - "3 Pi “x if eth - i i ® (a) Who are you friends with? (b) Who do you play with? @ B e 2 @ -@ yO eo @ 6 ~ | = 38 ~ Bh, rte B e e ‘ a e 8 e ° | e a ne @ 5 e 2 eo. a a e an 6m e 7 a @ e@ * B @ e a @ e / a = ¢ sg e° gs — He 5 @ @ ° a a Be a a ee e e@ aoa a G ° 5 e B B e a e 6 O B a e@ @ e ¢ ° ° a)

Figure 2 Network visualisations for the C&L school.

Regarding children as play isolates: the Semh school and the autism school recorded three (8% and 5%, re- spectively); the C&L school recorded eight (14%).

Negative network characteristics: Victimisation and perpetration

We observed considerably fewer negative than positive connections, but a higher number of nominations for experiencing victimisation than perpetration. On aver- age, pupils nominated fewer peers negatively (mean out- degree: 0.07—0.53), with 50%-88% isolates. Reciprocity and transitivity were generally low, although somewhat

(d) Who are you mean to? |

sporadic, likely due to the overall low number of nomi- nations made (the outdegree): 0%-32% of nominations were reciprocated, and 0%-75% of triplets were closed. Across networks, victimisation and perpetration were relatively evenly distributed across pupils (centralisa- tion) with a proportional range of 0.03—0.09, suggesting no clear hierarchy or specific targeting between peers in any of the three schools.

The autism school and the C&L school recorded similar rates of victimisation (M=0.53, Sd=1.02 and M=0.48, Sd=0.92, respectively) and reciprocation (0.32 and 0.21, respectively), with the Semh school reporting far lower rates of victimisation (M=0.26, Sd=0.64) and zero reciprocation. Reports of perpetration were very

rqvardde atp £q pouti9908 are sajanime yo :95n Jo sajna 10y ArexqrT atHE AoqEA\ Wo (stonsPUDD-Pur-sura\ Ios AapEAC rEEqHOUNOG/ScIY) StonNEPHRD pur stAIay, Xp 29 [9ZOZIOL/To] Ho ArexqET auHTUC ajLA, “(Pephurs 2p oSTUPY) UOISTAOLg TeHONEN suEyPO. YstHeds Kg Iziol'RLS-LOFT/T Tet -OL/Op/UOD-coqt-ArEaqe|aUNTHO'sTeAMofuaseU//:sdiy Wis Papeo{UNog “0 "Slssloft

asuaor] suum anv:

Social Networks In Special Schools

A Se a e ve As ° Oo oe e ao 7 te 7 re @ 6» o Be a eU66¢ © ee © ox» ° e w.e & a a e e ° a 1 @ a a a 8 @

a @ fe on e ° o @ @ a 8 ~. e ,. "sa e ° en e e e 8 _. ° a e ° as e © “ ° oe &, e ag @

(d) Who are you mean to?

Figure 3 Network visualisations for the autism school.

low across all three schools, with zero reciprocation in all cases.

Transitivity was highest for victimisation in the au- tism school (0.22) and for perpetration in the Semh school (0.75): in the Semh school it was more likely that bully perpetrators were connected. However, data were based on a very small number of connections. The number of isolates for the two bullying questions was high; approximately half of the pupils at the autism school and at the C&L school neither named a peer as having victimised them nor were named by a peer as causing victimisation (n= 32 (54%) and n=29 (50%), respectively). This increased to 62% of pupils (n= 24) at

the Semh school. These rates increased further when reporting perpetration, with the autism school report- ing 88% isolates (n=52), the Semh school reporting 79% isolates (n=31) and the C&L reporting 76% iso- lates (n=44).

Network bullying discrepancies

To investigate the agreement levels between bullying ac- cusations and confessions, we counted the number of matched and mis-matched nominations (see Table 2), fol- lowing the approach outlined in Tatum and Grund (2020).

fuaseujp'sdny t03y papeorun

uorstaorg reuoney aueaypop ystueds £4 1Z1OL'RLSS-LOFI/ Ttt Oi/oPAUOD Korte crexqqUT

vojpsdn) stonmpuo.y pure sux24, ap 225 “L9TOTIOL/Io] to Areaqy'y amu KajEAy “(PePHEES Bp Oui

sop Armage'y auHTUG) KaILAA Ho (StOR|PUED-puw-suAIOY/Ulo> K9A-ATEIGH

rqvandde axp kq pous9908 are sajone Yo :2sn 5

astta0r7 stot ane:

see the Terms and Conditions (htps://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; Oa articles are governed by the applicable Creative Commons License

14678578, 0, Downloaded from hitps://nasenjournals.onlinelibrary.wiley.com/doi/10.1111/1467-8578,70121 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on (01/10/20:

Badger er Al.

(I-09) uonstodoig

(suonosuuos Surreys

00°0 00°0 Slo tO 800 00°0 8r'0 1s0 1c0 6c0 bro £e0 suoneurmou) AArisuedy (I-09) uonstodoig

(sjenpratpur

puno.e suoneurmou

Jo wonenussuos)

£00 £00 S00 600 90°0 S00 cr0 £10 £10 £10 0c'0 Lv0 ToNestye.Us.) (I-09) uonsodoig (suonvurmou 000 00°0 00°0 ce0 1c0 00°0 cr0 790 Ir0 1s°0 09°0 8r'0 poqwoo.dioas) Ayoordioay, sjidnd jo raquinny cs by Te ce 6c ve € 8 € 0 0 z (suonevurwou 0.192) soye]OsT (as) W (6r'0) (waaid (sto) Lo0 = (I¥'0) 91.0 s1o (coDEsoO Coosro (90)97T0 §GrAZOE (Hid) 097 SDL8I GrOLTr (Ped 98'E (96'1) 69°7 suoneuron) 90139pnO % 6€ Ib s oyeurey (as) W (er D Tel (Ood $7 (rp ol sivak ul asy sjidnd jo raquinny 6S 8s 6£ Ozis Yiomion ushny T9O Hiwas mspny T9O His uspny Tvd Hingas uspny T9O Hiwas {0} UvoUT Noa Jib Oya {Nox 0} Urout St Oy Aa cum Avy nod op oy A Quiim spuolay nod o18 Oy Aa

‘(wistine pur Ppd ‘Hinas) sfooyps sary} oy} Jo yous Jj sonstyeys YIoMJou oaNdioseq = Atavl

Social Networks In Special Schools

Table 2. Numbers of accusations and confessions by school, to identify type 1 and type 2 discrepancies.

nasen|?

Possible interactions

Child 1 confesses

Child 1 does not confess

Semh 1482 Child 2 accuses 1 8 (type 1) Child 2 does not accuse 5 (type 2) 1468 C&L 3306 Child 2 accuses 4 25 (type 1) Child 2 does not accuse 6 (type 2) 3271 Autism 3422 Child 2 accuses 3 28 (type 1) Child 2 does not accuse 1 (type 2) 3390 Overall Child 2 accuses 8 61 (type 1) Child 2 does not accuse 12 (type 2) 8129 Note: Overall refers to the total bullying accusations, confessions, and discrepancies across schools. Table 3 Age and sex segregation across networks and schools. Expected chance level Observed: Friendship Observed: Victimisation Observed: (%) (%) Observed: Play (%) = (%) Perpetration (%) Age Semh 18 38 37 40 33 C&L 17 38 30 36 33 Autism 17 35 33 32 75 Sex Semh 90 95 93 100 100 C&L 51 68 77 57 44 Autism 52 58 60 65 50

Type | discrepancy: Accusation without confession

Definition: Child 2 accuses child 1 of bullying, but child 1 does not confess to bullying child 2. Across schools, there were 69 bullying accusations, of which only 8 (12%) were matched with a confession (88% type 1 discrepancy).

Semh school: There were nine instances of ‘child 2’ ac- cusing ‘child 1’ of bullying. However, in only one of those cases (11%) child 1 also confessed to having bullied child 2.

C&L school: There were 29 instances of ‘child 2’ ac- cusing ‘child 1’ of bullying. However, in only four of those cases (14%) did child 1 also confess to having bul- lied child 2.

Autism school: There were 31 instances of ‘child 2’ accusing ‘child 1’ of bullying. However, in only three of those cases (10%) did child 1 also confess to having bul- lied child 2.

Type 2 discrepancy: Confession without accusation

Definition: Child | confesses to bullying child 2, but child 2 does not accuse child 1 of bullying them. Across schools,

there were 20 bullying confessions, of which only 8 (40%) were matched with an accusation (60% type 2 discrepancy). Semh school: There were six instances of ‘child 1’ con- fessing to bullying ‘child 2’, However, in only one of those cases (17%) child 2 also accused child 1 of bullying them. C&L school: There were 10 instances of ‘child 1’ con- fessing to bullying ‘child 2’. However, in only four of those cases (40%) did child 2 also accuse child 1 of bullying them. Autism school: There were four instances of ‘child 1’ confessing to bullying ‘child 2’. In three of those cases (75%) ‘child 2’ also accused ‘child 1’ of bullying them.

Age and sex segregation

We compared the observed percentage of connections that were same age and same sex, for each school and network, to the percentage that would be expected based on chance given the age and sex composition of each school. The results are shown in Table 3.

For age, we observed considerable excess segrega- tion beyond expected chance levels: pupils were more likely to have both positive and negative connections with peers the same age as them than would be ex- pected by chance. Across schools, 17%-18% of con- nections would be expected to be the same age based

sd wwo3y papeoqumod “0 “SLS8L9FI

8 2 =

euoney aures4203 ypsmeds,

10] wo Kresge aumpuc Kayeay “(peprees 9p oxsaisn

ide ay Aq patzano8 ame sofonse yo :asn Jo saqns soy Areaqyy aurUG) KayrAA Ho (stoIpE

Badger er Al.

“| nasen on chance, whereas around 30%—40% were identified. Similar levels of age segregation were observed across schools and networks, with one exception: 75% of con- nections were same-age in the autism school bullying perpetration network, perhaps due to the low number of perpetration nominations.

We also observed considerable excess segregation be- yond expected chance levels for sex: pupils were more likely to have both positive and negative connections with peers the same sex as them than would be expected by chance. The expected percentage of same-sex con- nections was 51%—52% in the C&L and autism schools, but 90% in the Semh school. Despite these differences, segregation was consistently above chance levels across schools and networks, with two exceptions: the observed levels of same-sex connections were slightly lower than expected for the C&L and autism schools bullying per- petration networks, again perhaps due to the low num- ber of perpetration nominations. Overall, pupils showed an above-chance preference for forming same-age and same-sex connections, both positive and negative.

Discussion

As the first social network study conducted in the spe- cial school context, our findings provide unique insights into positive and negative peer connections among children with SEND in these important educational environments.

Social network data collection in special schools

We have shown that with slight adaptations, it is possible to successfully collect peer network data suitable for so- cial network analysis from pupils in special schools: out of a potential 191 eligible pupils, we successfully worked with 156 (82%), a participation rate comparable to main- stream school studies. The social stories, interview-style approach, and pre-populated drop-down option excel form worked well across all three schools and increased recording accuracy. The schools were fairly small and so further adjustments may need to be made if conducting future research in larger special schools.

Although very few parental opt-outs were received (n=1), 23 pupils did not assent on the day; two were too dysregulated and nine were persistently absent. The re- search team returned to each school on multiple days because in several cases pupils were not able to attend school full-time, for whole days, or had other sessions to participate in such as therapy or personalised educa- tional support. It is also important to note that partici- pation rate varied between schools. The Semh school had a much lower rate of participation (63%) compared to the C&L and autism schools (87% and 95%, respec- tively), with sixteen pupils not assenting and seven pupils

with persistent absence. This suggests that collecting data in Semh schools may prove more challenging.

Social networks in special schools

Previous literature suggested that pupils with SEND were less likely to have as many friendships compared to pupils without SEND, and that they would be more involved in bullying (Heiman, 2000; Pijl et al., 2008; Rose et al., 2011). In both cases, this would indicate that chil- dren with SEND are more likely than children without SEND to experience difficulties with peer relationships. Yet our findings show that in these three special schools, friendship, play, victimisation and perpetration networks largely mirrored the social networks observed in main- stream schools, suggesting that peer relationship diffi- culties may be less pronounced in these environments than might be expected based on previous research.

Positive networks (friendship and play)

The number of friendship connections ranged from 2.69 to 4.27 across schools (M=3.6), with a reciproca- tion rate range of 48%—60% (M=53%). These numbers are similar to previous research investigating friendship connections in mainstream schools: V6rés et al. (2019) found an average of 3.1 friendship nominations and a reciprocation rate of 45% in England. The pupils in our study did not have fewer reciprocated friendship connec- tions as perhaps previously suggested; in fact, they had slightly more on average. Our finding provides support for Holt et al. (2017) who noted that although the pupils with SEND in their sample (from both mainstream and special schools) had friendships, these were more preva- lent within the special school setting. Across all pupils, we found that only 1% were isolates in the friendship network, replicating work conducted by Bracegirdle et al. (2023) in mainstream schools.

Interestingly, in our study, there were fewer play con- nections than friendship connections, with all schools noting some play isolates (9%). Firstly, it is likely that pupils' perceptions of friendship and play are different. Some pupils may consider someone a friend if they sit next to them in class, but they would not choose to play with them. Whereas, at break time, they may choose to play with peers that they do not see as ‘situational friends’ in class, and/or instead view playmates as those with whom they play games or sports. Secondly, it is pos- sible that the word ‘play’ was inappropriate due to age. It is also possible that the word ‘play’ may have made pupils think of peers that they had played with that spe- cific day or week, whereas the word ‘friend’ may imply connections that are more stable over time.

For both friendship and play networks, we ob- served similar levels of transitivity to previous studies

8 2 = Z

10} Wo Kreaqr aun KaqeAg “(pEpHIES ap oUDN

w 204 meager] Buh] Kaiena Wo (SHORIPUOS-PuP-stUaY/Aios KapL"Aie

ide ayy Aq pawsano# ame safanue Vo ‘980 Jo

9 2 5

Social Networks In Special Schools

conducted in mainstream schools. On average, we ob- served 39% transitivity in our study, compared to 33% re- ported in a study of friendship networks in mainstream English schools (Bracegirdle et al., 2023). This indicates an equally high degree of tight-knit groups of peers in both mainstream and special schools, meaning that pos- itive connections are likely to form between two people who already share a common connection.

We also observed high levels of age and sex segrega- tion. This finding mirrors previous research conducted in mainstream schools that has observed pupils forming relationships with those who are similar (McPherson et al., 2001; Smith et al., 2014). The findings concerning age segregation are particularly of interest. Age segre- gation is not commonly studied in mainstream schools because pupils are typically divided into same-age class- rooms and grades, which form the network boundary. In many special school contexts, pupils mix more with peers across age groups, as schools are typically smaller and class membership can be based on a mixture of ed- ucational need, suitability and age. Our findings show evidence of age segregation despite opportunities to in- teract across ages.

Negative networks (victimisation and perpetration)

Again, contrary to what previous literature suggested, our findings show similar rates of victimisation and bullying across our three special schools, compared to mainstream schools. The number of victimisation connections ranged from 0.26 to 0.53 across schools (M=0.42), with a reciprocation rate range of 0%-32% (M= 18%), and 50%-—62% of pupils were isolates in the victimisation networks. This closely matches bullying data from British mainstream schools; for example, Bracegirdle et al. (2023) reported a range of 0.23-0.66 (M=0.43) connections across networks, with recipro- cation rates of 0%—12%, and 54%-73% isolates. Our findings suggest similarly low levels of victimisation among pupils in special schools, with the majority of pupils not involved in victimisation in any way. This observation differs from previous studies of bullying that have used alternative methodological approaches, which have suggested that bullying is much more com- mon among children with SEND, with rates up to 69% (Rose et al., 2011).

Of the four relationships, we observed the lowest number of connections for bullying perpetration, with an average of 0.07—-0.15 nominations and zero recipro- cation. This result is unsurprising as literature using al- ternative methodological approaches has also identified bullying perpetration as the lowest reported bullying role (Badger et al., 2023; Bowes et al., 2024). Interestingly, across schools, we observed 69 bullying accusations, of which only 12% were matched with a confession, and we

nasen |" observed 20 bullying confessions, of which 40% were matched with an accusation. Compared to mainstream schools, our type | discrepancy rate was slightly higher (74%, Tatum & Grund, 2020; 85%, Tolsma et al., 2013), whereas our type 2 discrepancy rate was lower (82%, Tatum & Grund, 2020; 78%, Tolsma et al., 2013). Overall, however, what pupils perceive or report as either bullying perpetration or victimisation are, in most cases, weakly matched.

Finally, we identified high levels of age and sex seg- regation for victimisation and perpetration, mirroring previous research conducted in mainstream schools: negative and positive social networks are divided along sociodemographic lines (Bracegirdle et al., 2023).

Comparisons to mainstream schools

Despite pupils' potential social and/or communica- tion difficulties and the contextual differences between special and mainstream schools, we observed largely similar findings to studies conducted in mainstream schools where the majority of pupils do not have SEND. This contrasts with previous research using alternative methodological approaches to Sna that have suggested children with SEND have comparatively fewer friends and experience more victimisation (Heiman, 2000; Pijl et al., 2008; Rose et al., 2011). Perhaps in environ- ments specifically structured to support pupils' needs and confidence, with specialist teachers and similar peers, pupils feel more able to form positive social connections.

Similarities and differences between school type

The social networks across all three schools showed similarly low levels of centralisation, indicating that the number of connections was relatively evenly dis- tributed across pupils within each school. However, we observed considerably more friendship and play connections, coupled with higher levels of reciprocity and transitivity overall, in the C&L and autism schools compared to the Semh school. These findings sug- gest a more closely connected peer social environment in the C&L and autism schools. Yet we also observed higher levels of victimisation in the C&L and autism schools. Thus, both positive and negative social con- nections appeared less common in the Semh school environment. In this school there were higher levels of persistent absence and more instances of pupils re- quiring personalised timetables, resulting in less class- room- and playground-based time with peers to form or sustain social connections.

The key difference between the autism and C&L school was the number of play isolates (5% and 14%, re- spectively). Perhaps the pupils at the autism school are

reuoney atesys09 ysturds £4 Izlol'SLSS-LOFL/Tl Tt Ot/topuOD:Kopracre3q

10} Wo Kreaqr aun KaqeAg “(pEpHIES ap oUDN

w 204 meager] Buh] Kaiena Wo (SHORIPUOS-PuP-stUaY/Aios KapL"Aie

ide ayy Aq pawsano# ame safanue Vo ‘980 Jo

9 2 5

Badger er Al.

=| nasen more socially inclusive, a finding that contradicts pre- vious work suggesting autistic children find it most dif- ficult to make and sustain friendships and playmates. It is possible that when autistic children are in an environ- ment specifically structured to support their needs and increase their confidence, they feel more able to trial and initiate friendships. However, it is unknown whether this school also embedded more explicit social and interac- tion support in their curriculum, which may have influ- enced these results.

Limitations and future work

First, it would be preferable to apply our research proce- dure in special and mainstream schools simultaneously to enable a more direct comparison. Second, this study provides a cross-sectional view of three schools, one of which (the Semh school) did not reach the desired response rate of 70%-—80% (Grosser & Borgatti, 2013). With only three schools, it remains unknown whether the observed differences result from differences in spe- cialism or other factors that could have played a part in the structure and tightness of the networks, includ- ing participation rate. Future research should study a larger number of special schools and consider other specialisms, in addition to tracking the social networks longitudinally to study potential differences in relation- ship stability across school types (Ladd, 1990). Future work could also consider applying Mixed Methods Sna (Mmsna; see Froehlich et al., 2020) to allow for a deeper understanding of the meaning and variations of relationships.

Conclusion

We have shown that the social network approach can be used with children with SEND, and with those in special schools. Overall, and contrary to what much literature would suggest, our findings show that the social net- works in special schools largely mirror those in main- stream schools: children with SEND experience similar numbers of friendships as those without SEND when in a safe and understanding environment. They also experi- ence similar patterns and rates of bullying involvement. However, we also identified key differences between special school types, with children in the Semh school experiencing fewer social connections overall, both posi- tive and negative, compared to those in the C&L and the autism schools. This research opens the pathway for future longitudinal social network analyses of so- cial relationships in special schools and their impacts on mental health, peer influence and the movement or sta- bility of friendships. Understanding social networks in special schools also provides important information for researchers, teachers and practitioners designing social

skills programmes, anti-bullying programmes, or per- sonal, social and emotional development programmes for pupils in the special school context.

Author Contributions

Ariel Lindorff: Funding acquisition; methodology; writ- ing — review and editing. Lucy Robinson: Formal analy- sis; investigation; methodology; project administration; software; writing — review and editing. Chloe Bracegirdle: Conceptualization; data curation; formal analysis; fund- ing acquisition; methodology; software; validation; visu- alization; writing — original draft; writing — review and editing. Kimberly Winardi: Formal analysis; investiga- tion; methodology; writing — review and editing. Julia R. Badger: Conceptualization; data curation; funding acquisition; investigation; methodology; project admin- istration; supervision; validation; writing — review and editing; writing — original draft.

Funding Information This work was funded by the John Fell Fund, University of Oxford [grant number 0015442].

Conflict Of Interest Statement The authors declare no competing interests.

Data Availability Statement

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

Ethics Statement

This study has been reviewed and given favourable opin- ion by the University of Oxford's Central University Research Ethics Committee (Curec): Reference num- ber R95252/RE001.

Orcid

Julia R. Badger ® https://orcid. org/0000-0003-3060-1934

Lucy Robinson ® https://orcid.org/0009-0004-4535-0191 Kimberly Winardi ® https://orcid. org/0009-0003-1923-2362

Ariel Lindorff ® https://orcid.org/0000-0001-6147-6289 Chloe Bracegirdle ® https://orcid. org/0000-0002-9936-2735

References

Badger, J.R., Zaneva, M., Hastings, R., Broome, M.R., Hayes, R., Patterson, P. et al. (2023) Associations between school-level dis- advantage, bullying involvement and children's mental health. Children, 10, 1852. Available from: https://doi.org/10.3390/child ren10121852

Bagwell, C.L. & Bukowski, W.M. (2018) Friendship in childhood and adolescence: features, effects, and processes. In: Bukowski, W.M., Laursen, B. & Rubin, K.H. (Eds.) Handbook of peer in- teractions, relationships, and groups. New York: Guilford Press, pp. 371-390.

8 2 = Z

10} Wo Kreaqr aun KaqeAg “(pEpHIES ap oUDN

w 204 meager] Buh] Kaiena Wo (SHORIPUOS-PuP-stUaY/Aios KapL"Aie

ide ayy Aq pawsano# ame safanue Vo ‘980 Jo

9 2 5

Social Networks In Special Schools

Block, P. (2015) Reciprocity, transitivity, and the mysterious three- cycle. Social Networks, 40, 163-173. Available from: https://doi. org/10.1016/j.socnet.2014.10.005

Bowes, L., Babu, M., Badger, J.R., Broome, M.R., Cannings-John, R., Clarkson, S. et al. (2024) The effects and costs of an anti- bullying programme (KiVa) in UK primary schools: a mul- ticentre cluster randomised controlled trial. Psychological Medicine, 54, 4362-4373. Available from: https://doi.org/10. 1017/S0033291724002666

Bracegirdle, C., Jonsson, J.O. & Spiegler, O. (2023) Neither friend no foe: ethnic segregation in school social networks. Socius: Sociological Research for a Dynamic World, 9, 1-3. Available from: https://doi-org/10.1177/23780231231214956

British Educational Research Association [Bera]. (2024) Ethical guidelines for educational research, 5th edition. Bera. Available from: www.bera.ac.uk/publication/ethicalguidelines-for-educa tional-research-2024

Campbell, M., Straatmann, V.S., Lai, E.T.C., Potier, J., Pinto Pereira, S.M., Wickham, S.L. et al. (2019) Understanding social inequali- ties in children being bullied: UK millennium cohort study find- ings. PLoS One, 14(5), 1-15. Available from: https://doi.org/10. 1371 /journal.pone.0217162

Chamberlain, B., Kasari, C. & Rotheram-Fuller, E. (2007) Involvement or isolation? The social networks of children with autism in reg- ular classrooms. Journal of Autism and Developmental Disorders, 37, 230-242. Available from: https://doi.org/10.1007/s1080 3-006-0164-4

Christina, S., Magson, N., Kakar, V. & Rapee, R. (2021) The bidirec- tional relationships between peer victimization and internaliz- ing problems in school-aged children: an updated systematic re- view and meta-analysis. Clinical Psychology Review, 85, 101979. Available from: https://doi.org/10.1016/j.cpr.2021.101979

Department for Education. (2026) SEND assessment guidance and resources. Available from: https://help-for-early-years-providers. education.gov.uk/support-for-practitioners/send-assessment

Froehlich, D.E., Van Waes, S. & Schafer, H. (2020) Linking quanti- tative and qualitative network approaches: a review of mixed methods social network analysis in education research. Review of Research in Education, 44(1), 244-268. Available from: https:// doi.org/10.3102/0091732X 20903311

Gonzalez, M.C., Herrmann, H.J., Kertész, J. & Vicsek, T. (2007) Community structure and ethnic preferences in school friendship networks. Physica A: Statistical Mechanics and its Applications, 379, 307-316. Available from: https://doi.org/10.1016/j.physa. 2007.01.002

Gov.UK. (2025) Special educational needs in England. Available from: _ https://explore-education-statistics.service.gov.uk/find- statistics/education-health-and-care-plans/2025 [Accessed 3rd September 2025].

Granovetter, M.S. (1973) The strength of weak ties? American Journal of Sociology, 78, 1360-1380.

Grosser, T.J. & Borgatti, S.P. (2013) Network theory/social network analysis. In: McGee, R.J. & Warms, R.L. (Eds.) Theory in so- cial and cultural anthropology: an encyclopaedia. Los Angels, LA: Sage Publications, pp. 595-597.

Heider, F. (1946) Attitudes and cognitive organization. Journal of Psychology, 21, 107-112. Available from: https://doi.org/10.1080/ 00223980.1946.9917275,

Heiman, T. (2000) Friendship quality among children in three ed- ucational settings. Journal of Intellectual and Developmental Disability, 25, 1-12. Available from: https://doi.org/10.1080/13269 7800112749

Holt, L., Bowlby, S. & Lea, J. (2017) Everyone knows me I sort of like move about: the friendships and encounters of young peo- ple with special educational needs in different school settings. Environment and Planning A: Economy and Space, 49, 1361-1378. Available from: https://doi.org/10.1177/0308518X 17696317

nasen 1

eos

Ladd, G.W. (1990) Having friends, keeping friends, making friends, and being liked by peers in the classroom: predictors of chil- dren's early school adjustment? Child Development, 61, 1081— 1100. Available from: https://doi.org/10.2307/1130877

Locke, J., Ishijima, E.H., Kasari, C. & London, N. (2010) Loneliness, friendship quality and the social networks of adolescents with high-functioning autism in an inclusive school setting. Journal of Research in Special Educational Needs, 10, 74-81. Available from: https://doi.org/10.1111/j.1471-3802.2010.01148.x

McPherson, M., Smith-Lovin, L. & Cook, J.M. (2001) Birds of a feather: homophily in social networks. Annual Review of Sociology, 27, 415-444. Available from: https://doi.org/10.1146/ annurev.soc.27.1.415

Moody, J. (2001) Race, school integration, and friendship segrega- tion in America. American Journal of Sociology, 107, 679-716. Available from: https://doi.org/10.1086/338954

Pakaslathi, L. & Keltikangas-Jarvinrn, L. (2001) Peer-attributed prosocial behavior among aggressive/preferred, aggressive/non- preferred, non-aggressive/preferred and non-aggressive/non- preferred adolescents. Personality and Individual Differences, 30, 903-916. Available from: https://doi.org/10.1016/S0191-8869(00) 00082-9

Pijl, S.J., Frostad, P. & Flem, A. (2008) The social position of pupils with special needs in regular schools. Scandinavian Journal of Educational Research, 52(4), 387-405. Available from: https://doi. org/10.1080/00313830802184558

Pinto, C., Baines, E. & Bakopoulou, I. (2019) The peer relations of pupils with special educational needs in mainstream primary schools: the importance of meaningful contact and interaction with peers. British Journal of Educational Psychology, 89, 818— 837. Available from: https://doi.org/10.1111/bjep.12262

Rivera, M-T., Soderstrom, S.B. & Uzzi, B. (2010) Dynamics of dyads in social networks: assortative, relational, and proximity mech- anisms. Annual Review of Sociology, 36, 91-115. Available from: https://doi.org/10.1146/annurev.soc.34.040507.134743

Rose, C.A., Monda- Amaya, L.E. & Espelage, D.L. (2011) Bullying perpetration and victimization in special education: a review of the literature. Remedial and Special Education, 32, 114-130. Available from: https://doi.org/10.1177/0741932510361247

Salmivalli, C., Huttunen, A. & Lagerspetz, K.M.J. (1997) Peer networks and bullying in schools. Scandinavian Journal of Psychology, 38, 305-312. Available from: https://doi.org/10.1111/ 1467-9450.00040

Schoeler, T., Duncan, L., Cecil, C.M., Ploubidis, G.B. & Pingault, J.B. (2018) Quasi-experimental evidence on short- and long- term consequences of bullying victimization: a meta-analysis. Psychological Bulletin, 144, 1229-1246. Available from: https:// doi.org/10.1037/bul0000171

Smith, S., Maas, I. & van Tubergen, F. (2014) Ethnic ingroup friend- ships in schools: testing the by-product hypothesis in England, Germany, The Netherlands and Sweden. Social Networks, 39, 33-45. Available from: https://doi-org/10.1016/j.socnet.2014.04. 003

Tatum, T.G. & Grund, T.U. (2020) Accusation and confession discrep- ancies in bullying: dual-perspective networks and individual- level attributes. Social Networks, 60, 61-70. Available from: https://doi.org/10.1016/j.socnet.2019.09.001

Tolsma, J., van Deurzen, I., Stark, T. & Veenstra, R. (2013) Who is bullying whom in ethnically diverse primary schools? Exploring links between bullying, ethnicity, and ethnic diversity in Dutch primary schools. Social Networks, 35, 51-61. Available from: https://doi.org/10.1016/j.socnet.2012.12.002

Van Harmelen, A.-L., Kievit, R., Ioannidis, K., Neufeld, S., Jones, P., Bullmore, E. et al. (2017) Adolescent friendships predict later resilient functioning across psychosocial domains in a healthy community cohort. Psychological Medicine, 47, 2312-2322. Available from: https://doi.org/10.1017/s0033291717000836

5 : z

mage aUNTUO AaqEAK “(Peiups ap Outs

ipuED pu sua, a 295 “9;

& £ : : z $ £

asuaor] suourtno.y aaneaiy ajqporidde axp &q powsoso8 ame sajonre Vo

“I nasen Badger er aL.

Volk, A.A., Dane, A.V. & Marini, Z.A. (2014) What is bullying? A theoretical redefinition. Developmental Review, 34, 327-343. Available from: https://doi.org/10.1016/j.dr.2014.09.001

Vordos, A., Block, P. & Boda, Z. (2019) Limits to inferring status from

How to cite this article: Badger, J.R., Robinson, L., Winardi, K., Lindorff, A. & Bracegirdle, C. (2026)

friendship relations. Social Networks, 59, 77-97. Available from: Social networks in special schools: Friendships, https://doi.org/10.1016/j.socnet.2019.05.007 bullying and play among pupils with special

World Health Organization. (2020) Youth violence. Available from: educational needs and disabilities. British Journal Betps:wwse-who-int/news-room/fack-sheets/detail/youth of Special Education, 00, 1-14. Available from: -v .

Zablotsky, B., Bradshaw, C.P., Anderson, C.M. & Law, P.A. (2013) https://doi.org/10.1111/1467-8578.70121

Risk factors for bullying among children with autism spectrum disorders. Autism, 18, 419-427. Available from: https://doi.org/10. 1177/1362361313477920

9 Zz z

mage aUNTUO AaqEAK “(Peiups ap Outs

ipuED pu sua, a 295 “9;

& £ : : z $ £

& 5 F 5 E 5