Intermediaries and cross-examination resilience in children: The development of a novel experimental methodology

Experimental studies examining child 'witnesses' under cross-examination typically rely on researchers questioning children using a 'barrister's script'. In the current research, experienced barristers used a defence statement from a mock perpetrator (who committed a theft 11 months earlier) to challenge typically developing chil-dren's evidence under cross-examination. We also assessed whether Registered Intermediaries (RIs), trained professionals who facilitate communication between vulnerable witnesses and members of the justice system, help children reduce compliance with misleading cross-examination suggestions. Results demonstrated that children (6 – 11 years) complied with barristers' challenges to a high degree: 94% agreed with at least one of the barristers' seven false suggestions. However, when assisted by an RI, children were significantly less compliant with barrister challenges. These findings, and additional analyses of the nature of child responses and barrister questions, provide novel exploratory evidence for the beneficial role of RIs in temper-ing the adverse effects of cross-examination style questioning for children.


| INTRODUCTION
In adversarial justice systems, such as England and Wales, child witnesses in criminal trials provide their evidence-in-chief (direct evidence) via video-recorded Achieving Best Evidence investigative interviews (Ministry of Justice, 2011). Subsequently, they may be questioned on this evidence by the opposing counsel ('cross-examination'), who has an interest in undermining this evidence. This can mean that witnesses, "having first been questioned by someone who wants them to say one thing… are then cross-examined by another person who wants to make them say the opposite" (Spencer, 2012, p. 1). Here, we report the development of a novel experimental methodology to investigate cross-examination performance in typically developing children. We also assess whether providing child witnesses with a 'Registered Intermediary' (RI; a trained professional who facilitates communication between vulnerable witnesses and members of the justice system, Ministry of Justice, 2020a) improves the quality of children's evidence, by reducing compliance with barrister challenges about false information.
Recommendations of the Pigot Commiteee (Home Office, 1989) led to legislation in England and Wales that enabled, with the agreement of the court, vulnerable and intimidated witnesses to benefit from 'special measures' (Youth Justice and Criminal Evidence Act, 1999). These included: screens (preventing the witness from seeing the defendant); live links (enabling the witness to give evidence during the trial from outside the court room via a televised link); the removal of wigs and gowns (by judges and barristers); pre-recorded video evidence-in-chief and cross-examination; use of aids for communication (enabling questions or answers to be communicated to or from the witness); and examination of the witness assisted by an RI. Although most of these recommendations have since been fully implemented in England and Wales (the jurisdiction relevant to the current study), live-link cross-examinations were largely retained. 1 Improving the quality and reliability of children's evidence under cross-examination is an urgent international priority given serious concerns about how child witnesses are treated in criminal courts (e.g., Andrews et al., 2015a;Spencer, 2012;Zajac et al., 2012). Studies of court transcripts (e.g., Australia, England, New Zealand, Scotland, USA) highlight that large proportions of questions posed to children during cross-examination are inconsistent with best practice guidelines and developmental level, with heavy reliance on closed, optionposing, suggestive (leading), repeated, and complex questions (e.g., Andrews et al., 2015a;Andrews et al., 2015b;Andrews & Lamb, 2016;Evans et al., 2009;Hanna et al., 2012;Hanna & Henderson, 2018;Klemfuss et al., 2014;Zajac & Cannan, 2009). Suggestive questions are particularly problematic, as the likelihood of errors increases with their use (Lamb et al., 2011). Such questions "should only be used as a last resort and only when necessary (e.g., to immediately safeguard a person)" (Bull, 2010, p. 9), yet they are commonly recommended to advocates to maintain control of the discourse (Hanna et al., 2012). This illustrates the conflict between the aims of cross-examination (to test evidence) and best practice guidelines (to elicit evidence) (Zajac et al., 2012). Indeed, some have called crossexamination "a virtual 'how not to' guide to investigative interviewing" (Henderson, 2002, p. 279), directly violating methods that promote completeness and accuracy (Zajac et al., 2012) and exploiting children's vulnerabilities . Almost 90% of witnesses under 11-years do not understand questions they are asked at court (Plotnikoff & Woolfson, 2009). Further, almost 95% of crossexamination transcripts of child sexual abuse cases reveal inconsistencies, largely between what is said in police interviews relative to subsequent cross-examination (Pichler et al., 2020). Worryingly, a comparative study of child sexual abuse case transcripts in Australia found no improvements in the format of questions used over the past 60 years (leading questions still predominated), with more questions asked, which were more likely to be complex (Zajac et al., 2018). Empirical studies of cross-examinations support these findings, noting that high numbers of children change their responses following questioning. In children of 4-11 years, 70%-98% changed at least one aspect of their testimony when challenged (e.g., Bettenay et al., 2014;Righarts et al., 2015;. Most previous empirical studies employed researchers challenging witnesses by asking scripted crossexamination questions, although occasionally trainee legal professionals have been used (e.g., Bettenay et al., 2014). Yet, it is more realistic to allow barristers free reign to tackle cross-examinations in the way they see fit. In the present study, an unscripted approach was used to assess cross-examination compliance in children, enabling barristers to adapt according to the way a child responded, and to press points more emphatically if they were making headway, which is not possible using a script.
The study also investigated whether one of the special measures, the Witness Intermediary Scheme (available in England and Wales since 2004), would help reduce child witnesses' compliance with barrister challenges about false information. The role of RIs is wideranging but includes assessing the communication abilities of vulnerable witnesses and offering impartial and specific advice on posing best practice questions by accommodating each individual child's language and communication needs. The aim is to facilitate communication between the child and relevant professionals to ensure it is complete, coherent and accurate (Collins & Krahenbuhl, 2020;Cooper & Wurzel, 2014;Krahenbuhl, 2019). Several other international jurisdictions (e.g., Northern Ireland, New Zealand, Norway, New South Wales, Australia) have adopted intermediary schemes, although details of the schemes vary (see Cooper & Mattison, 2017;Cooper & Wurzel, 2014;Taggart, 2021). Feedback on the RI scheme has been generally positive (Collins & Krahenbuhl, 2020;Ministry of Justice, 2020a;Plotnikoff & Woolfson, 2015), and mock juror studies suggest that the presence of an RI does not have a negative impact on perceptions of child witnesses (e.g., Krahenbuhl, 2019). However, further empirical evidence in relation to RI use during mock crossexaminations is needed and the current study offers exploratory evidence in this regard.
The current study forms part of a broader research programme examining child witness performance during all stages of a mock criminal investigation: initial statements (Henry, Messer, et al., 2017); investigative interviews ; identification line-ups (Wilcock et al., 2018(Wilcock et al., , 2019; and cross-examinations (presented here). Children viewed a staged event involving a minor mock crime (in which one man 'stole' another man's phone or keys) and were cross-examined on this evidence approximately 11 months after undergoing initial investigative interviews (representing close to the average delay of 8 months for a case to go to trial in England and Wales at the time of the study; Plotnikoff & Woolfson, 2012). Qualified, experienced barristers took on the role of the defence barrister and were presented with a defence statement with which to question the children, allowing the barrister to adopt an unscripted approach.
The first primary research question was whether, and to what extent, children would comply with the barrister's challenges on seven elements of false information in the statement. A second primary research question considered whether providing child witnesses with RI assistance reduced compliance with the barrister's challenges on this false information (a proportion of our sample was assisted by a fully qualified, experienced RI at all stages of giving formal evidence).
Given the lack of previous empirical evidence, predictions were tentative. We hypothesised that: (1) children would comply to a large degree with barrister challenges on false information; and (2) a beneficial effect of RI assistance on compliance with false information on cross-examination challenges would emerge, as RIs facilitate communication, for example, rephrasing questions in a developmentally appropriate manner in line with an individualised communication assessment. Two subsidiary research questions were also addressed: (3) in RI assisted cross-examinations, would children's responses show less compliance (and more resistance) to challenges on false information?; and (4) in the RI condition would barristers change the style and nature of questions in line with the recommendations given for questioning (based on each child's communication assessment and according to best practice for interviewing young children)? We tentatively predicted that children in the RI condition would be less likely to comply with, and more likely to resist, challenges on false information; and that barristers would ask more questions in the RI condition consistent with best practice. The broader research programme included a control interview condition (Best-Practice) and two other interview conditions (Sketch-Reinstatement of Context and Verbal Labels). We did not expect the two other interview conditions to differ from the Best-Practice condition in terms of cross-examination resistance or nature of responses/questions.

| Participants
A total of 202 typically developing children were recruited from mainstream primary schools in London and the Southeast of England, but three were excluded: one had a full-scale IQ in the intellectual disability range; and two were unavailable for the investigative interview (see Henry, Crane, et al., 2017, for further details). Of the remaining 199 children, 177 (84 boys, 93 girls) were available for crossexamination 11 months later (range 8-13 months). At this stage, one further child (a girl) was excluded because she did not respond to any cross-examination questions. The remaining 176 children ranged in age from 6 years 7 months to 11 years 3 months (mean = 8 years 6 months, SD = 1 year 2 months) at the time of the initial investigative interview; and 7 years 7 months to 12 years 3 months (mean = 9 years 5 months, SD = 1 year 2 months) at the crossexamination stage. See Table 1 for details.

| Materials and procedure
As described, this research was part of a wider project exploring the performance of child witnesses across different stages of the criminal justice process (children on the autism spectrum were included, but we were unable to cross-examine enough children to ensure reliable findings with this group). Of relevance to the current paper, were three phases.  Henry, Messer, et al., 2017, for further information).

| Phase 2: Investigative interviews
Approximately 1 week later, children took part in one of four types of investigative interview (see , for further information).

Best-practice
Based on Achieving Best Evidence principles (Ministry of Justice, 2011), this interview comprised seven key phases: (1) greet and personalise the interview; (2) rapport building (chatting to the child about areas of interest); (3) truth and lies exercise (e.g., determining whether the child correctly responds to a statement along the lines of 'that lady is wearing a blue jumper' when it is red); (4) explain the purpose of the interview; (5) free recall

Registered intermediary (RI)
Here, children were supported by one of two experienced, practising RIs. Prior to the interview, the RI individually assessed each child and there was a meeting between the RI and each interviewer to discuss recommendations for the interview and to flag any individual needs. RIs advised the interviewers to follow the protocol for the Best-Practice interview, with some adaptations (e.g., simplifying the verbal instructions given to the children, and recommending the use of visual aids that were provided by the RIs). At all times, the RI was present to facilitate communication between the child and the interviewer. As the interviewer proceeded through the Best-Practice interview protocol, the RI intervened when appropriate to facilitate effective communication (verbally or by suggesting the use of suitable props).

Verbal labels
This followed the procedure for the Best-Practice interview except that, following phase 5 (free recall), witnesses received 'tell me more' prompts in relation to four key areas (adapted from Brown & Pipe, 2003): (1) , 2015). This is standard practice for witnesses in advance of cross-examination within courts in England and Wales. Therefore, as in real-life, cross-examination performance may draw upon original memories of the event and recent memories of the refreshed interview. The researcher visited the child to explain that, in the next day or so, they would be speaking to a barrister who would ask them some questions about the staged event they previously saw.
The researcher explained that the child would be listening to the audio of their interview, 4 to remind them of the event and what they had said. After refreshing of the evidence, the researcher again reminded the child about the forthcoming cross-examination.
A team of six barristers was recruited for the cross-examinations, comprising four men and two women. Five were currently practising barristers, whilst one was no longer practising but had their own legal business. Barristers had between 5-21 years of criminal law experience (mean = 15.2 years).

Cross-examination: A new methodological approach
For the cross-examination, a 'defence statement' was developed for each version of the staged event, which the barristers were asked to put to the children. This created a more realistic situation in which the barrister was representing a defendant in relation to a charge of theft.
The defence statement (and the cross-examination protocol) was developed with the advice and guidance of an experienced barrister.
The first two items in the statement included correct information designed to set the scene, establish rapport with the child witness, and make them feel at ease. The remaining points contained an element of untruthfulness (except for points 6 and 7, which were included so children did not feel that they were disagreeing with all the points the barrister was raising). Table 2 provides a sample defence statement for one version of the event.
Barristers were asked to challenge the child on all seven of the 'false' points (e.g., "I think you've got a little bit mixed up because it wasn't the phone that Adam put in his pocket, it was the keys, wasn't it?") a maximum of four times (a decision, in consultation with one of the barristers, to avoid ethical concerns). As there was variability in this (based on barrister judgement), scores only reflect whether a child complied immediately, following challenge/s, or not at all. If the child complied with the challenge on first time of asking, they received a resistance score of 0; if they complied with a challenge on the second or subsequent time of asking, they received a resistance score of 1; and if they did not comply at all, they received a maximum resistance score of 2. Average resistance scores on each of the seven false points could range from 0-2, with higher scores indicating higher cross-examination resilience (i.e., lower compliance with false statements). On a few occasions, barristers judged that it was not necessary to pose all challenges to the children. In real life, barristers make judgements about how much/little to press a witness and do not take a fixed approach, so the present study aimed to reflect this. Therefore, mean resistance scores were calculated for each child based on the total number of challenges given.
We were careful to code the child's original recall of information pertaining to each of the seven false points (taken from the investigative interview), so this score could be controlled in the analyses. These 'memory trace' scores were allocated for full (3), moderate (2), partial (1) or no (0) knowledge about six of the false points in terms of degree of information recalled in the investigative interview. For one other point (false item 5), this was a complete confabulation about something that did not happen at all in the event, therefore, a score of 0 was allocated for all children because it was not possible to code this item in terms of original recall of information (maximum memory trace score = 18: see Table 1 for mean memory trace scores and Appendix S1 for full details of the coding scheme). Fifteen percent of the transcripts were independently coded by a second rater for memory trace scores and intra-class correlations for information pertaining to each of the challenges ranged from .89 to 1.00, indicating excellent inter-rater reliability.

Cross-examination: The study protocol
One special measure available to support vulnerable witnesses in courts in England and Wales is the 'live link'. The child is not present in the courtroom with the barristers, judge or jury, but is in a separate room. Those in the courtroom see the child via a television screen, and the child can see the judge or barrister on his/her screen. To mimic this, cross-examinations were performed using video conferencing software (Skype). A female researcher was in a room with the child at their school and partially took on the role of 'judge'. We could not entirely replicate the judge role as we had no facility for the child to view the judge only via the screen-and for ethical reasons the researcher had to be with the child-so this aspect of the study must be viewed as approximate to real-life. There was a brief 'ground rules hearing' between the judge and the barrister prior to each individual cross-examination (with or without an RI) where the judge explained any important considerations to the barrister (e.g., age of child, any additional needs they had). As a prelude to the cross-examination, the judge explained to the child that they: (1) needed to tell the truthmust not guess or make anything up; (2) could say that they 'do not Once the child and barrister were introduced, they listened to the child's audio of their investigative interview together, so everyone could hear it (barristers were provided with a transcript of the children's testimony, as well as basic demographic information, in advance of the cross-examination, to enable them to prepare their questions; in real-life, they would have access to the child's evidencein-chief in advance of the refreshing of the evidence). The barrister then began questioning the child, with the only stipulations being that they were to cover all points on the defence statement (unless the child appeared to show any signs of distress), and that-for ethical reasons-they were not to excessively challenge the child on their testimony (no more than four challenges per point).
At three time points (before, during and after the cross-examination), children were presented with a 10-point visual analogue rating scale. This enabled us to monitor how worried or anxious the children were (1 = no anxiety; 10 = high anxiety) and to offer additional support or reassurance if their responses highlighted that they were affected by the cross-examinations. Note that these anxiety ratings were not study variables but introduced for ethical reasons. Most children were not highly anxious at any point. Before the cross-examination, 7 children (4%) had scores at the top end of the anxiety scale (8, 9, 10); during the cross-examination this figure was 9 children (5%); after the cross-examination nearly all (171 children, 97%) had the lowest anxiety scores of 1, 2 or 3 (and the remaining 5 children had moderate scores of 4, 5 or 6). Cross-examinations were, on average, 8.56 min long (SD = 2.24 min, range 3.53-16.25 min).

Cross-examination protocol: The RI condition
The protocol for the cross-examinations was the same across three interview conditions (Best-Practice, Sketch-RC and Verbal Labels), but there were some differences for the RI condition. As per T A B L E 2 Sample defence statement from one of the two versions of the event (including the 'truth' and the seven 'false' statements) Points from the defence statement The 'ground truth'-From the event 1 One morning last year, Max and I visited a school to give a talk about the Victorians to the children and their teachers True 2 Max was wearing a blue top and has short brown hair. I was wearing a grey top and had long blond hair tied back in a ponytail.
True 3 When we arrived, a woman helped us by setting up the video camera at the back which recorded the talk.
False item 1-Adam set up the video camera. There was no woman involved in the event. 4 We told the children some rules that Victorian children had to obey, for instance, we said that boys must learn needlework False item 2-whilst the children were told about rules, this specific example is incorrect-the children were told that girls (not boys) had to learn needlework.

5
We showed the children a slate and Max showed them how to write the letters of the alphabet on it with chalk.
False item 3-the children were shown a slate, but Max wrote a sum on the slate (not the alphabet). 6 Max is very forgetful and during the talk he asked the children to remind him not to forget his phone at the end of the talk.
True 7 Max then put his phone on the chair in the hall. True 8 Max says that I stole his phone by taking it and putting it in my pocket-I did not do this. Max's phone was on the chair the whole time. I did not go near the chair at any time during or after the talk.
False item 4-Adam did take Max's phone and put it in his pocket.

9
I did borrow Max's keys during the talk and put them in my pocket.
False item 5-there were no keys involved in the staged event.
10 At the end of the talk, Max forgot his coat. False item 6-Max forgot his jumper (which he spoke about at the start of the talk).

11
When Max forgot his coat, I had to go back to get it. False item 7-Max (not Adam) returned after he had left, to collect the forgotten item.
Note: Whilst the other version of the event was very similar, points 4-11 on the defence statement differed: for example, there were slightly different names (Mark and Alex) for the key actors; children saw the theft of a set of keys, but the barrister had to put to them that it was, in fact, a phone; and the children were told that boys had to learn technical drawing (with the barristers suggesting to them that this was girls).
recommendations for best practice in England and Wales at the time of the study (Registered Intermediary Procedural Guidance Manual, Ministry of Justice, 2015), children received RI assistance both at their initial interview and again at cross-examination. Of the 33 children in the RI condition, 18 were assisted by the same RI at both stages, which is also recommended best practice, and 15 had a different RI at cross-examination (although using exactly the same protocol). In real cases there is also likely to be some variability in whether the same RI is available for both stages. RI assistance involved the following: Prior to the cross-examination, all children were re-assessed by the RI to ensure that information about the child's communication needs (originally collected 8-13 months previously) was up-to-date and accurate.
This re-assessment took place at least a week before the crossexamination and consisted of: (1) re-establishing rapport with the children; (2) explaining what would happen in the cross-examination; (3) checking the children could say they 'do not know' or 'cannot remember', and could state whether the barrister (adult) was wrong or right; (4) checking the children could respond to questions beginning with, for example, 'when' or 'how'; and (5)

Coding child responses and barrister questions
Children's responses were coded into mutually exclusive categories reflecting whether they complied, resisted, did not respond, responded with an open question, or sought clarification (see Table 3). When a child responded with an acknowledgement (e.g., 'okay'), this was not coded as a response to the question. If the child said they were not sure, this did not mean they had complied: children were instructed to say 'do not know' if this was the case, so they were resisting the barrister's attempts to get them to agree with them.
Barrister questions were coded into one of seven overarching mutually exclusive primary categories (see Table 4 for details). All questions (as well as non-content-based utterances which were given the code 'other') were coded separately, even if they occurred, sequentially, e.g. "That's really helpful, thank you very much happened in Victorian times? (code=invitation closed, true)" would attract three codes as indicated. Barrister questions were additionally coded for each instance of 17 other secondary features (see Table 5), which were not mutually exclusive categories, that is, a question could challenge credibility as well as contain a tag. The coding systems were developed by looking at guidance on questioning available at the time

Reliability of coding
To establish coder agreement, 10% of scripts were coded independently by a second coder. Overall percentage agreement was 91% (range 86%-100%) for the child codes, 89% (range 82%-92%) for the barrister primary codes and 88% (range 81%-100%) for the barrister secondary codes, all of which represented moderately high agreement.

| Control measures
Around the time that the children took part in Phases 1 and 2 of the study, several cognitive measures (intelligence, language, memory, attention) were administered to ensure factors that may affect eyewitness recall and cross-examination were controlled or matched between interview groups (see Table 1 for differences between conditions that were controlled for statistically).

Intelligence
Two subtests (Vocabulary and Matrix Reasoning) of the second edition of the Wechsler Abbreviated Scale of Intelligence (WASI-II; Wechsler & Zhou, 2011) were used to provide an assessment of intellectual ability and to establish suitability for entry into the study.

Language
The

Attention
The Test of Everyday Attention for Children (Tea-Ch; Manly et al., 1999) was used to assess a range of relevant attention skills: selective/focused attention (the 'Sky Search' subtest); sustained attention (the 'Score!' subtest); and sustained-divided attention (the 'Sky Search Dual Task' subtest).

| General procedure
Ethical approval was obtained from the relevant university Research Ethics Committee. Prior to participation, written consent was obtained from parents, and children also gave their own written assent to participate. At the start of Phase 1, children viewed the "Lovely, thank you so much B."; "Can I ask you some questions about that because that's really helpful?" staged event and immediately took part in the brief interviews (see Henry, Messer, et al., 2017). Phase 2, investigative interviews (see  and identification line-ups (see Wilcock et al., 2018), took place around 1 week later. Cognitive testing also took place around this time, which was split over several sessions to fit in with school timetables and to ensure children remained engaged with tasks. Phase 3, the cross-examinations, took place 8-13 months (Mean = 11.06 months, SD 1.69 months) after viewing the staged event. As some variability in this delay emerged across conditions (see Table 1) due to timing of school holidays and availability of RIs/barristers, we controlled for delay in the primary statistical analyses. All children were refreshed on their evidence in one session with the researcher, before the researcher returned at least 1 day later to conduct the cross-examination with the barrister. Children in the RI condition were re-assessed in a session prior to the refreshing of their evidence (on a different, earlier day). The RI was always present at the cross-examination and, beforehand, used a visual aid to explain to the child that they should only say what really happened, that if the barrister got something wrong, they could tell them, and equally that it was OK to say that the barrister 'got it right'. In addition, the children were told, using the visual aid, that it was OK to say 'I don't know', 'I can't remember', or 'I don't understand'.

| RESULTS
The key outcome measures for the primary research questions con-

Repetition
Repeating the same question, even if interspersed by others "Did Alex take the keys?" A: "No". "Did Alex take the keys?"

Confirmation
The advocate confirms the answer the child has given, in a best practice way-a permissible and gentle way of checking evidence "I want to make sure I understand what you said…" "so they showed you the slate but they did not do any writing, is that what you are saying?"

Clarification
The advocate checks that the answer the child has given is what was intended "You nodded, so is that a yes, brilliant, thank you very much."

Social influence of another person
The barrister suggests that 'someone else' told them that what the child has said happened did not really happen "Alex told me he did not take the keys" Questions asking the witness if they remember what they said on a previous occasion are particularly frowned upon "Do you remember any other adults in the room?" "Can you remember that?" Lying Directly accuses the witness of lying Note: an example is not given as there were no examples of accusing the child of lying in the current study.

Signpost
Explaining or signposting changes of subject (includes references to original evidence, e.g., "in your interview, you said that...") "Now we are going to talk about the other man, the man with the long hair called Adam." Praise Thanking or commending the child in an encouraging way "That's brilliant, thank you for that. I've only got one more thing to ask you…" Filler Irrelevant questions "The men who came to your school, were they funny?" Name The advocate uses the child's name "That's really helpful, you have got a very good memory here N."

Reassurance
The advocate provides reassurance that the child is doing okay "That's okay, not to worry, so you cannot help me with who set it up if you do not remember." cross-examination challenges pertaining to false elements from the defence statement; and (2) whether RI assistance during crossexaminations reduced children's compliance with these challenges on false information. Hierarchical multiple regression was used to examine whether cross-examination resistance scores on the seven false information challenges differed between children in the RI condition versus other conditions (note that we had no reason to expect cross-examination differences for the Sketch-RC and Verbal Labels conditions as they involved adaptations to investigative interview protocols). At step 1, three background variables showing differences between interview conditions (see Table 1 for details), namely age at cross-examination, IQ, and Verbal Memory, were controlled (BPVS scores also differed between interview conditions, but IQ and BPVS scores were highly correlated, r = .66, so only IQ was controlled). Three further control variables included: memory trace scores (concerning relevant information pertaining to the false information challenges) as children in the RI condition had higher memory trace scores (they had benefitted from RI intervention at the investigative interview stage) ; event version (A or B); and length of delay before cross-examination (this differed across condition-see Table 1). At step 2, three dummy-coded interview condition variables were included to test for differences between conditions in cross-examination resistance. Best-Practice was the reference (control) group to which the other three conditions were compared: RI, Sketch-RC and Verbal Labels. The dependent variable was average cross-examination resistance score (see Table 6). With nine predictor variables in total, Green (1991)   The results were identical in all respects, except that memory trace score at Step 1 just missed significance (p = .055). 5  Table 6 includes mean proportions of the seven types of responses.

| Children's responses
Proportional data were not all normally distributed, so Kruskal-Wallis tests were used to explore whether there were differences between interview conditions for each type of response, with a Bonferroni adjusted significance level of p < .007 (for seven tests).
Two analyses were of relevance to predictions as follows. For Complies (

| Barrister questions
A second subsidiary research question concerned whether, in the RI condition, the barristers' questions might be more consistent with best practice guidance for cross examination or re-examination.  tions did not differ significantly (p = .08). Given these differences, further analyses on barrister questions were performed using T A B L E 7 Summary of the multiple regression predicting average cross-examination resistance Step B SE B β p Step Cross-exam delay (months) .04 .03 .14 .10 Step 2 proportional scores: the total number of questions in each questiontype category were divided by the total number of barrister questions asked per child. These proportional data were not all normally distributed, so Kruskal-Wallis tests were used to explore whether there were interview condition differences on each question type, with a Bonferroni adjusted significance level of p < .007 (for seven tests).
Bonferroni corrected follow-up paired comparisons were used to explore any differences between interview conditions. Invitation Closed (false information) questions (p = .10) and Other questions (p = .03) showed no significant interview condition differences. Table 9 includes breakdowns of barrister questions into 17 secondary features. These are presented as proportions (i.e., divided by the total number of barrister questions), but will not add up to one given the categories are not mutually exclusive (any question could be classified in one or more ways). (Note: no instances of the barrister saying the child was 'lying' were found; similarly, mean proportions for use of idiom were less than 1%; so these data were excluded.) These proportional data were not all normally distributed, so Kruskal-Wallis tests were used to explore interview condition differences for each question feature, with a Bonferroni adjusted significance level of p < .003 (for 15 tests). Bonferroni corrected follow-up paired comparisons were used to explore any differences between interview conditions.

Invitation
Eight secondary question features showed significant interview condition differences.

| DISCUSSION
In this paper, a novel experimental methodology for the crossexamination of vulnerable child witnesses has been presented. Experienced barristers questioned children based on a 'defence statement' containing seven false elements, without recourse to a 'script' (as is typically used in experimental research on cross-examination). As predicted, children complied with barristers' challenges on this false information to a high degree: 94% of children complied with at least one cross-examination challenge on false information, consistent with previous experimental studies using scripted questioning in which compliance rates ranged between 70% and 98% (cf. Bettenay et al., 2014;Righarts et al., 2015; cross-examination is a reliable method for obtaining best evidence from child witnesses, given that lawyers try to 'persuade children to change details in their accounts, often by exploiting their developmental limitations' (Andrews & Lamb, 2016, p. 953).
We also tested, in an exploratory way, whether RI assistance, ideally have an intermediary assessment as, no matter how advanced they appear, their language comprehension is likely to be less than that of an adult witness" (paragraph 98, p. 60). For typical children, RIs also help improve volume of recall in interviews and accuracy of identification in video line-ups Plotnikoff & Woolfson, 2012;Wilcock et al., 2018). Overall, providing A final subsidiary research question concerned whether barristers would ask questions more aligned with best practice recommendations in the RI condition. In support of this, barristers asked proportionally more Invitation Open questions in the RI condition. Whilst these have been associated with inconsistencies (due to the longer answers they elicit) (Pichler et al., 2020), they are in accord with best practice (Ministry of Justice, 2011), are least likely to lead the witness , and are highly valued by practitioners (Magnusson et al., 2020). Invitation Open questions were, nevertheless, relatively rare, as reported in real cases (e.g., Andrews & Lamb, 2016;Pichler et al., 2020;Zajac et al., 2018). Rates here ranged from 4% to 5% in Pre-trial guidance aimed at children may also help because practice sessions for responding to cross-examination style questions on an unrelated topic can significantly improve children's overall accuracy during a subsequent cross-examination interview (Irvine et al., 2016;Righarts et al., 2013), provided it is given close to the interview date . Future research could investigate a combination of RI assistance and timely pre-trial preparation (perhaps delivered as part of the RI assessment), as combining these interventions may further improve the quality of children's cross-examination evidence.
One area the study was unable to illuminate was whether the RI assistance impacted on the child's responses, the barrister's questioning technique, or both. We are also uncertain about the mechanisms and exact points through which RI assistance operated, but it is important to note that the overarching role of the RI is to support the child's communication needs (e.g., simplifying instructions, using visual aids) and impact the barrister's questioning to ensure it is appropriate. All of this should help the witness more easily understand what others are saying so that they can communicate better. Further research could unpick the important mechanisms underpinning the interplay between children's responses and barristers' questions.
The very nature of cross-examination requires some fluidity in questioning and a good advocate will always be influenced by the child's responses. The exception to this would be to use a rigid script of questions (which is necessary in some extreme cases, but not generally). Otherwise, the barrister can be flexible and adapt in response to the child's answers. This was one of the advantages to the present novel approach to assessing cross-examination empirically, which has, to our knowledge, not been addressed in previous empirical work.
There are some limitations to the study that should be acknowledged. One is that the findings are applicable only to defence barristers, as different lines of questioning may be applied by prosecution barristers (Denne et al., 2020). Another is that children in the RI condition, as per best practice guidance (Ministry of Justice, 2015), had already received RI assistance during previous phases of the mock criminal investigation: this was given at the investigative interview stage (which also included an identification line-up). Therefore, the current conclusions can only be applied to children who have had RI assistance throughout a criminal investigation which, in practice, is not always the case (RIs may sometimes only brought in at trial stage, although this is not recommended). A related issue was that children in the RI condition remembered more about the initial witnessed event, as RI assistance was effective in increasing the volume of accurate recall at investigative interview . This meant that children in the RI condition started their cross-examination with a recall advantage. We mitigated this by controlling for how well the child had recalled key facts about the false information in the defence statement (memory trace scores). Although memory trace was not a significant predictor of cross-examination resistance in the full regression (and many children did not score highly on this measure), future research could match on initial memory of the staged event before instigating cross-examinations in groups with and without RI assistance. This method would mean that no children could be included who had previously undergone an investigative interview assisted by an RI, but such a method would provide evidence about the effectiveness of RI assistance brought in only at the trial stage.
Further limitations are as follows. We used a mild minor crime event that took place in a familiar environment (the children's school), so were unable to replicate the anxiety, unfamiliarity and potential trauma of a real court case, which limits generalisation of the findings.
Children were seen by friendly and supportive researchers, and the barristers were also approachable and experienced-they were, partly, chosen on the basis of having previous experience in cross-examining children (for ethical reasons)-again, this might not be so in real-life.
Our ground rules hearings for non-RI children were also brief, and more recent guidance now recommends they are included as 'good practice' for all young witnesses (Judicial College, 2018, revisions 2020. Finally, the length of the cross-examinations, for ethical reasons was short (average 8.56 min) compared to real cases (reported in England and Wales as between 45 min and 3 h, Baverstock, 2016). However,  reported much shorter video-recorded cross-examinations (16 min) in a pilot trial of this special measure in England, and with new advocate training and guidance, cross-examinations are likely to be more limited in length (e.g., Judicial College, 2018). Similarly, although studies of court transcripts in Scotland, California and New Zealand have emphasised the large numbers of questions (ranging from 160 to 500) posed to children by prosecutors and defence lawyers (Andrews & Lamb, 2016;Andrews et al., 2015a;Klemfluss et al., 2014;Zajac & Cannan, 2009), the number of questions posed during pilot videorecorded cross-and direct-examinations in Henderson et al.'s (2019) study was lower (average = 92). Thus, although the current crossexaminations contained fewer questions (average = 61), the overall numbers of questions may be more aligned with the newer prerecorded cross-examinations in England. Given that long and complex cross-examinations will likely lead to fatigue, worsening the quality of evidence (e.g., Zajac et al., 2018), changes that encourage shorter questioning should be advantageous.

| CONCLUSION
The current study was the first to use a more ecologically valid defence statement as the basis for unscripted cross-examinations.
Using this novel method, we found that children complied with a very high number of barrister challenges on false information. However, we also found exploratory evidence that RI assistance reduced children's compliance with barristers' cross-examination challenges on false information. This could be, in part, because the barristers asked questions that were somewhat more aligned with best practice recommendations in the RI condition. These findings extend previous research on the utility of RIs during investigations (evidence-gathering interviews and identification line-ups). They provide additional evidence of the importance of using RIs to ensure typically developing young children can give accurate testimony during the final investigative phase (cross-examination).