Data collection included two distinct groups of participants. (1) Data collected from 202 typically developing children and 72 children with autism, all in the age range 6-11 years. (Note that one child with autism and one child with typical development had an IQ below 70.) Data collected from the children included standardised and non-standardised tests for a range of cognitive skills (IQ, language, memory, attention, suggestibility, anxiety) as well as data from the four experimental phases of the research: brief initial interviews; full investigative interviews; identification line-ups; and cross-examinations (note that not all children participated in this final phase). (2) Data collected from three samples of jury eligible adults (ages 18-69, n=260 in total) concerning their credibility ratings of interviews (and some cross-examinations) of selected children with and without autism. Children with Autism Spectrum Disorder (ASD) display a characteristic pattern of memory and social communication difficulties that may call into question their abilities as witnesses. The current research examined how children with and without ASD fared during all the different stages of a criminal investigation, and considered the best ways of eliciting reliable evidence. There were four research questions: 1. How do children with ASD fare during the different stages of a criminal investigation (initial questioning, an investigative interview, when identifying perpetrators, and during cross-examination) relative to typical children? 2. What can be done to improve the amount of information that children recall, without making them more prone to errors; for example, is the assistance of a Registered Intermediary - a professional communications specialist who assists vulnerable individuals within the criminal justice system - useful? 3. How do the general public – who may be evaluating evidence within a jury - perceive child witnesses with and without ASD? 4. Can we predict how well children will perform as witnesses? For example, is it the case that children with good language, attention or memory skills, are more likely to provide full and accurate statements.

(1). Data from child 'witnesses'. Samples of typical children (6- to 11-year-olds) were recruited from five mainstream schools in Greater London. Samples of children with autism (6- to 11-year-olds) were recruited from mainstream schools in Greater London, special schools in Greater London, and via parent support networks in Essex and Kent. Children were assessed with standardised cognitive tests (and some unstandardized tasks). Procedure for the mock criminal investigation. 1. Children viewed an event at school (either live or via a video) about school in Victorian times involving two male adult actors, one of whom committed a minor theft. 2. Children received a brief interview about this event on the same day - akin to a first response police officer taking an initial statement. 3. One week later children received a full investigative interview and viewed two identification line-ups (relating to the two male actors). Prior to this, children were semi-randomly allocated into four interview conditions: a standard police interview; a Verbal Labels interview (additional support from distinct questions about people, actions/objects, setting and conversations); a Sketch Reinstatement of Context interview (using a drawing technique to re-orientate the child to the context of the event); or a Registered Intermediary supported interview (a fully-qualified experienced intermediary supported and facilitated the child's communication throughout the interview and identification line-up). 4. 9-12 months later, available children in the sample were cross-examined by experienced barristers over Skype to mimic the 'live link' used in court. A defence statement was prepared and the barrister challenged the child on seven key points from this statement, attempting to get them to agree to erroneous information. Scoring for the child witnesses data. All standardised tests were scored according to the relevant manuals. Interviews (Brief and Investigative) were recorded, transcribed and scored according to clear and robustly derived criteria based on the number of 'units of correct information' recalled. These data were additionally scored for the types of details recalled (e.g. numbers of person, action, object, setting, conversation, and general details). Numbers of incorrect details (information that was inaccurate) and confabulations (information that did not occur in the event) were also recorded. Line-up choices were scored as correct identifications, incorrect identifications, incorrect rejections, or correct rejections. Mock cross-examinations were scored based on how many times, from a maximum of seven points in the 'defence statement', the child ceded to challenges posed by the barrister (higher scores for those who were more resistant to the barristers' challenges). We also recorded the extent to which children had provided information about these seven points in their investigative interviews (memory trace score - higher scores for those who provided fuller details). (2). Data from 'mock jurors'. Jury eligible adults (18-69 years) were recruited via convenience sampling from universities, workplaces and personal contacts. These participants rated selected interviews/cross-examinations from our child witnesses dataset on a series of 7-point scales, reflecting how they appraised the credibility of that individual witness. Study 1 (n=120) - ratings were made of interviews with children who had autism. Two children with autism provided interviews and half of the participants rated each video. Further, participants were divided into three groups, randomly assigned to receive: (1) no information about the child's autism diagnosis; (2) information that the child had a diagnosis of autism; or (3) information that the child had a diagnosis of autism together with a summary of the key characteristics of autism. Study 2 (n=60) – ratings were made of two videos of children with autism; one child received a standard police interview and the other child was supported by a registered intermediary. Participants rated one video and were randomly allocated to one of the two video conditions. Study 3 (n=80) - ratings were made of interviews and cross-examinations of four typically developing children: two of these children were supported by a registered intermediary and two of these children received a standard police interview. In each pair of children, one child was 'higher' ability and one child was 'lower' ability and these pairs were matched across interview conditions. Participants each rated one interview and were randomly allocated to one of the four conditions.