Quantifying Language Experience in Bilingual and Trilingual children, 2020-2022

The data in this repository were collected in France, the Netherlands and the UK between 2020 and 2022 to inform the validation of the Q-BEx questionnaire (). Children between the ages of 5 and 9 were tested individually to assess their proficiency in the societal language (i.e., French, Dutch or English), as well as their non-verbal intelligence and working memory. One of their parents or caregivers also filled in the full version of the Q-BEx questionnaire. The respository includes data from 299 children (FR: n=78, NL: n=117, UK: n=104), although some measures are not available for all children.
In France, children were recruited in ordinary schools and in private clinics for Speech & Language Therapy (37 children were recruited via clinics). The consent form for schools asked parents if the child had previous or current SLT, and the reason. In the Netherlands, recruitment took place in schools and via social media advertisement. Language disorder (reported by parent/teacher/remedial teacher) was an exclusionary criterion. In the UK, all children were recruited in schools; no exclusionary criteria were applied, and not SLT information was collected.
Language experience data*
Language experience data was collected using the full version of the Q-BEx questionnaire (), which was completed by one of the child’s parents or caregivers. This includes all the following modules (except that Language mixing wasn’t included in France):
- Background (languages the child is exposed to, adults and children the child lives with
- Risk factors (early language milestones, early parental concerns)
- Language exposure and use (current and cumulative estimates; onset of exposure to each language)
- Estimates of proficiency in each language (listening, speaking, reading, writing)
- Richness of experience in each language (activities, diversity of interlocutors, parental education)
- Language mixing
- Attitudes
The questionnaire was administered either in the societal language (French, Dutch or English) or in one of the child’s home languages (Arabic, Dutch, English, French, German, Italian, Polish, Romanian, Russian, Spanish, Turkish). The choice of administration language was constrained by the translated versions available at the time of testing. The translation protocol used to create the versions in different languages can be found at .
Direct outcome measures
We collected measures of language and cognitive outcomes during individual, face-to-face sessions with each child (two sessions per child, lasting approximatively 45 minutes each). Most of the testing was done in the child’s school. In France, the children recruited via speech & language therapy clinics were tested in the clinic. In the UK and the NL, some testing sessions took place in a different location (e.g., university premises), and on rare occasions online via Zoom.
Language proficiency
Outcomes in the societal language (i.e., Dutch, English, or French) include phonology, morphosyntax and vocabulary.
Phonology
Phonological competence was assessed with the LITMUS Quasi-Universal Non-Word Repetition task. See dos Santos, C., and Ferré, S. (2016) “A Nonword Repetition Task to Assess Bilingual Children’s Phonology”. Language Acquisition 41: 1–14.
Morphosyntax
Morphosyntax outcomes were assessed with the LITMUS test in each societal language. See Marinis, T. and S. Armon-Lotem (2015). “Sentence Repetition. Methods for assessing multilingual children: disentangling bilingualism from Language Impairment.” in S. Armon-Lotem, J. de Jong and N. Meir, Methods for assessing multilingual children: disentangling bilingualism from Language Impairment. Amsterdam: Multilingual Matters).
The English and Dutch versions included 30 items. The French version included 16 items. We created two blocks in the English and Dutch versions so that the first block be comparable to the 16-item French version. Subsequent analyses demonstrated that there was no block effect in EN and NL.
For each child, we report three overall scores: Identical Repetitions, Target Repetitions, and Grammatical Attempts. These overall scores correspond to the mean across items (n=30 in English and Dutch; n=16 in French), excluding NAs (i.e., the mean is calculated on the items for which the child did provide a response).
Vocabulary
Vocabulary breadth was measured in the UK with the British Picture Vocabulary Scale (BPVS), in France with the the Échelle de vocabulaire en images Peabody (EVIP), and in the Netherlands with the Peabody Picture Vocabulary Test (PPVT).
Vocabulary depth was measured with the Word Classes component of the Clinical Evaluation of Language Fundamentals CELF-V (in its Dutch, English, or French version).
Cognitive measures
The tasks used to evaluate cognitive skills were administered in the child’s societal language.
Memory
Short-term memory was assessed through Forward Digit Recall; working memory was assessed through Backward Digit Recall.
Most children were tested using the digit span protocols described in Hill et al (2021), implemented in Psychopy (to allow randomisation of the digit sequences and facilitate the acquisition of detailed data). Children were presented with sequences of numbers (through headphones) and asked to repeat these numbers either in the same order (in the FDR task) or in reverse order (in the BDR task). The length of the sequence increased by one digit after 4 trials, starting with 3 digits in the first block of the FDR task, and 2 digits in the first block of the BDR task. The maximum sequence length was 6 digits in the FDR task, and 5 digits in the BDR task.

Children recruited via the Speech and Language Therapy clinics in France (n=37) experienced more difficulty with this task, so it was decided to use the WISC-V protocol instead for these children, as it included a discontinuation rule.

To allow comparison across groups, we created a WISC-like score for the data collected via the Hill et al (2021) protocol. This consisted in the digit span for which the child had at least one fully accurate response (i.e., all the digits in the span, in the right order) for at least one out of the first two trials for that span (as the WISC-V protocol only features 2 trials per digit span).

FDR_overall and BDR_overall correspond to the total accuracy scores, as per the Hill et al (2021) protocol.
FDS_Q and BDS_Q correspond to either the WISC-V score (for children recruited via clinics) or the WISC-like score created as explained above, depending on which protocol the child was tested with. All children have a FDR_Q and a BDR_Q score in the dataset.

Non-verbal intelligence
The matrices task from either the Wechsler Intelligence Scale for Children–Fifth Edition (WISC–V) or the Wechsler Preschool & Primary Scale of Intelligence - Fourth UK Edition (WPPSI-IV) was used to measure non-verbal intelligence - depending on the age of the child: children below the age of 6 were tested with the WPPSI.

Depending on the age of the child at the time of testing, we used the WPPSI protocol (for children younger than 6 years of age, n= 77) or the WISC-V protocol (for all the other children).

本仓库中的数据于2020年至2022年期间在法国、荷兰和英国收集，旨在为Q-BEx问卷（）的验证提供依据。研究人员对5至9岁的儿童进行了个体测试，以评估其社会语言（即法语、荷兰语或英语）的熟练程度，以及非言语智力（non-verbal intelligence）和工作记忆（working memory）。儿童的父母或照料者之一也填写了Q-BEx问卷的完整版。该仓库包含299名儿童的数据（法国：n=78，荷兰：n=117，英国：n=104），不过部分测量指标并非所有儿童都有记录。

在法国，研究人员从普通学校和私人言语语言治疗（Speech & Language Therapy, SLT）诊所招募儿童（其中37名来自诊所）。学校的知情同意书（consent form）询问父母孩子是否有过或正在接受SLT治疗及其原因。在荷兰，招募渠道包括学校和社交媒体广告；语言障碍（由父母、教师或辅导教师报告）是排除标准。在英国，所有儿童均从学校招募，未设置排除标准，且未收集SLT相关信息。

语言经验数据*
语言经验数据通过Q-BEx问卷完整版（）收集，由儿童父母或照料者之一填写。该问卷包含以下所有模块（法国样本未纳入语言混合模块）：
- 语言背景（儿童接触的语言类型、共同生活的成人及同伴）
- 风险因素（早期语言发展里程碑、父母早期担忧）
- 语言接触与使用（当前及累积接触量、各语言接触起始时间）
- 各语言熟练程度评估（听、说、读、写）
- 各语言经验丰富度（活动类型、对话者多样性、父母教育水平）
- 语言混合
- 态度
问卷以社会语言（法语、荷兰语或英语）或儿童的母语之一（阿拉伯语、荷兰语、英语、法语、德语、意大利语、波兰语、罗马尼亚语、俄语、西班牙语、土耳其语）施测。施测语言的选择受当时可用的翻译版本限制。不同语言版本的翻译方案可参见 。

直接结果测量
研究人员通过与每名儿童进行一对一面对面测试（每名儿童两次测试，每次约45分钟）收集语言及认知结果数据。大部分测试在儿童所在学校进行。在法国，从言语语言治疗诊所招募的儿童在诊所接受测试；在英国和荷兰，部分测试在其他地点（如大学场地）进行，极少数情况下通过Zoom在线完成。

语言熟练程度
社会语言（即荷兰语、英语或法语）的测试结果包括语音学、形态句法和词汇三个方面。

语音学
语音能力通过LITMUS准通用非词重复任务（LITMUS Quasi-Universal Non-Word Repetition task）评估。参见dos Santos, C. 与 Ferré, S. (2016) 的「评估双语儿童语音能力的非词重复任务」，载于《语言习得》（Language Acquisition）第41期：1-14页。

形态句法
形态句法结果通过各社会语言版本的LITMUS测试评估。参见Marinis, T. 与 S. Armon-Lotem (2015) 的「句子重复：评估多语言儿童的方法——区分双语能力与语言障碍」，收录于S. Armon-Lotem、J. de Jong与N. Meir主编的《评估多语言儿童的方法——区分双语能力与语言障碍》（Methods for assessing multilingual children: disentangling bilingualism from Language Impairment），阿姆斯特丹：Multilingual Matters出版社。英语和荷兰语版本包含30个项目，法语版本包含16个项目。研究人员在英语和荷兰语版本中设置了两个模块，以便第一个模块可与法语版本的16个项目进行比较。后续分析表明，英语和荷兰语版本中不存在模块效应。每名儿童的报告包含三个总分：完全重复分（Identical Repetitions）、目标重复分（Target Repetitions）和语法尝试分（Grammatical Attempts）。这些总分是所有项目的平均值（英语和荷兰语n=30，法语n=16），不包含缺失值（即仅计算儿童有回应的项目的平均值）。

词汇
词汇广度在英国通过英国图片词汇量表（British Picture Vocabulary Scale, BPVS）测量，在法国通过皮博迪图像词汇量表法语版（Échelle de vocabulaire en images Peabody, EVIP）测量，在荷兰通过皮博迪图片词汇测试（Peabody Picture Vocabulary Test, PPVT）测量。词汇深度通过《临床语言基础评估第五版》（Clinical Evaluation of Language Fundamentals, CELF-V）的词汇分类模块（荷兰语、英语或法语版本）测量。

认知测量
认知能力评估任务以儿童的社会语言施测。

记忆
短时记忆通过顺向数字回忆（Forward Digit Recall, FDR）评估；工作记忆通过逆向数字回忆（Backward Digit Recall, BDR）评估。大多数儿童采用Hill等人（2021）描述的数字广度方案测试，该方案通过Psychopy软件实现（可随机化数字序列并便于收集详细数据）。儿童通过耳机听取数字序列，然后按相同顺序（FDR任务）或相反顺序（BDR任务）重复。序列长度每4次试验增加1位：FDR任务第一模块从3位开始，BDR任务第一模块从2位开始。FDR任务的最大序列长度为6位，BDR任务为5位。法国从言语语言治疗诊所招募的儿童（n=37）完成该任务时难度较大，因此决定改用韦氏儿童智力量表第五版（Wechsler Intelligence Scale for Children-Fifth Edition, WISC-V）方案，因其包含终止规则。为便于组间比较，研究人员为采用Hill等人（2021）方案收集的数据创建了类WISC分数。该分数定义为：儿童在某一数字广度的前两次试验中，至少有一次完全正确回应（即数字广度中的所有数字按正确顺序）的数字广度值（因WISC-V方案每个数字广度仅包含2次试验）。FDR_overall和BDR_overall是根据Hill等人（2021）方案得到的总准确率分数。FDR_Q和BDR_Q分数取决于儿童所采用的测试方案：若为诊所招募儿童，则为WISC-V分数；否则为上述类WISC分数。数据集中所有儿童均有FDR_Q和BDR_Q分数。

非言语智力
非言语智力通过韦氏儿童智力量表第五版（WISC-V）或韦氏学前及初小儿童智力量表英国第四版（Wechsler Preschool & Primary Scale of Intelligence-Fourth UK Edition, WPPSI-IV）的矩阵任务测量，具体取决于儿童年龄：6岁以下儿童采用WPPSI测试。根据儿童测试时的年龄，研究人员采用WPPSI方案（6岁以下儿童，n=77）或WISC-V方案（其他所有儿童）。