To the Editor,

Martin et al. contribute interesting observations on parent-infant interaction within the early prodrome of ASD. However, the inference they draw from their observations, namely that ‘insecure and insecure-resistant attachments are noteworthy precursors of later diagnosis’, is debatable.

The paper reports observations using the standard Strange Situation Procedure (SSP). This observes infant behaviour in a defined experimental condition, namely reunion with caregiver after a planned separation of a length intended to mobilise the infant's underlying ‘attachment behaviour system’, theorised to be largely quiescent in every day interaction unless mobilised by conditions of threat or loss. This then allows a specific inference from the behaviour to infant attachment dynamics. A corollary is that behaviours classified within the SSP will not necessarily relate to everyday naturalistic observations in a low stress situation; something supported generally by empirical research within attachment theory.

Within this SSP reunion procedure, Martin et al. observe high levels of infant emotional intensity and negative affect towards the parent; behaviours indicating an insecure resistant classification with attachment coding—and this accounts for the inference they make in their paper. The group incidentally do not find high levels of ‘disorganisation’ of attachment in the sample, something that has been typically associated with ASD in the past.

Their observations echo our own findings on infants at elevated familial risk of autism relative to low-risk controls (n = 91) made within the British Autism Study of Infant Siblings (BASIS) cohort (Wan et al., 2013). At 12–14 months, we found in these high-risk infants that lower positive affect and attentiveness to parent alongside less dyadic mutuality was specifically associated with an ASD diagnosis at 3 years. Importantly, these observations were made in an intentionally low-stress free play lab setting rather than in a high stress SSP paradigm, since our interest was to plot the evolution of naturalistic social communication between infant and parent from 8 months in both at risk and normative infants, as part of a transactional account of the early ASD prodrome (Green, 2019; Sameroff, 2009; Wan et al., 2019). Our inference from this and other longitudinal observations is that these interactive behaviours form part of the transactional precursor trajectory towards autism emergence in the first years; indeed are a rather sensitive marker of it.

Such findings of negative infant affective responses to parent associated with ASD but outwith the SSP context, introduce a key potential confound into the Martin et al. coding inference around attachment.

That standard attachment coding inferences can be confounded by altered neurodevelopment has been well-recognised and much discussed since the early studies of Capps et al. and the important 1999 SRCD Monograph that they reference. Most other studies in this area have subsequently adjusted attachment codings to exclude inferences from autism-specific behaviours—and at least this should have been addressed in their discussion.

Martin et al. confine themselves to reporting just two measures in their report (15 month SSP and 36 month ADOS outcome). To investigate the potential confound we point up here they could for instance have reported a 15 month measure of autism pre-symptoms (e.g., the Autism Observation Scale for Infants, AOSI)—a measure that would be standard in ‘babysibs’ studies of this kind. They do report a group comparison on the adjustment for attachment disorganisation advised by Pipp-Siegal et al., but this is not relevant to their reported insecure-resistant attachment coding.

Does all this matter? We have interestingly convergent early behavioural observations linked to later ASD diagnosis on similarly sampled risk-infants; but importantly in very different lab settings, designed for different inferences. The unique property of the SSP is to allow specific inferences about the attachment behavioural system. The fact however that the same behaviours are seen in low-stress interactions as a precursor to ASD outcomes, suggests not an attachment-specific inference but one related to emerging transactional trajectories of social communication development.

Yes! on the one hand this matters greatly. Inferences about disrupted attachments as a precursor of autism have a lot of historical salience in autism, stemming from the notorious (disproven) ‘refrigerator mother’ aetiological ideas from psychoanalytic theory. Developmental research has not supported an attachment origin for autism, which seems to be the inference they make.

On the other hand, both sets of observations relate to early parent infant interaction within emergent autism. Their suggestions for intervention strategies flowing from such observations, which target optimisation of such interaction, are indeed relevant. For instance, our iBASIS intervention for at-risk siblings in the first year was adapted from an original intervention aimed at improving attachment relationships; but in our hands the adaptation intentionally focused on dyadic social communication as a precursor of ASD rather than attachment; while still aiming to optimise dyadic social interaction and communication within early development. In an RCT with longitudinal follow-up, iBASIS intervention succeeded in producing significant sustained reduction in prodromal ASD symptom severity to 3 years (Green et al., 2017). Of course, such dyadic improvement may also benefit parent-child relational quality as well as child symptom severity, but this is a far cry from inferring that attachment insecurity is a causal precursor for autism, for which we would say there is no evidence.

致编辑，

马丁等人。对 ASD 早期前驱症状中的亲子互动做出了有趣的观察。然而，他们从观察中得出的推论，即“不安全和不安全的依恋是后来诊断的值得注意的前兆”，是值得商榷的。

该论文使用标准的奇怪情况程序 (SSP) 报告了观察结果。这观察了婴儿在定义的实验条件下的行为，即在计划分离一段旨在调动婴儿潜在的“依恋行为系统”的长度后与照顾者团聚，理论上在每天的互动中基本上是静止的，除非受到威胁或失去的条件的动员. 然后，这允许从行为到婴儿依恋动态的特定推断。一个推论是，在 SSP 中分类的行为不一定与低压力情况下的日常自然观察有关。依恋理论中的实证研究普遍支持的东西。

在这个 SSP 重聚程序中，Martin 等人。观察到婴儿的高情绪强度和对父母的负面影响；行为表明带有附件编码的不安全抵抗分类 - 这解释了他们在论文中做出的推论。该小组偶然在样本中没有发现高水平的依恋“混乱”，这在过去通常与 ASD 相关。

 他们的观察结果与我们在英国婴儿兄弟姐妹自闭症研究 (BASIS) 队列（Wan et al., 2013）中对自闭症家族风险相对于低风险对照组（n = 91）的婴儿的发现相呼应。在 12-14 个月时，我们发现在这些高危婴儿中，较低的积极影响和对父母的关注度以及较少的二元相互关系与 3 岁时的 ASD 诊断特别相关。重要的是，这些观察是在有意低压力的自由游戏实验室环境中进行的，而不是在高压力的 SSP 范式中进行的，因为我们的兴趣是绘制婴儿和父母之间从 8 个月以来处于危险和规范中的自然社会交流的演变婴儿，作为早期 ASD 前驱症（Green，2019 年；萨梅洛夫，2009；Wan 等人，2019 年）。我们从这个和其他纵向观察的推断是，这些互动行为构成了头几年自闭症出现的交易前兆轨迹的一部分。确实是它的一个相当敏感的标志。

这种婴儿对与 ASD 相关但与 SSP 相关的父母的负面情感反应的发现，给 Martin 等人带来了一个关键的潜在混淆。围绕附件的编码推断。

自从 Capps 等人的早期研究以来，这种标准的附件编码推断可能会被改变的神经发育所混淆，这一点已得到广泛认可和讨论。以及他们引用的重要的 1999 年 SRCD 专着。该领域的大多数其他研究随后调整了依恋编码，以排除自闭症特定行为的推论——至少这应该在他们的讨论中得到解决。

马丁等人。仅限于在报告中报告两项指标（15 个月的 SSP 和 36 个月的 ADOS 结果）。为了调查我们在这里指出的潜在混淆，例如，他们可以报告 15 个月的自闭症前兆症状测量（例如，婴儿自闭症观察量表，AOSI）——这是“婴儿”研究中的标准测量种类。他们确实报告了 Pipp-Siegal 等人建议的关于依恋混乱调整的组比较，但这与他们报告的不安全的依恋编码无关。

这一切重要吗？我们对类似采样的风险婴儿进行了与后来 ASD 诊断相关的有趣收敛的早期行为观察；但重要的是，在非常不同的实验室环境中，为不同的推理而设计。SSP 的独特属性是允许对依恋行为系统进行特定推断。然而，在低压力互动中将相同的行为视为 ASD 结果的前兆这一事实表明，这不是一种特定于依恋的推论，而是与社会传播发展的新兴交易轨迹有关的推论。

是的！一方面，这非常重要。关于被破坏的依恋是自闭症前兆的推论在自闭症中有很多历史意义，源于精神分析理论中臭名昭著的（被证明的）“冰箱母亲”的病因学观点。发展研究不支持自闭症的依恋起源，这似乎是他们做出的推论。

另一方面，两组观察结果都与出现自闭症的早期父母婴儿互动有关。他们从这些观察中得出的干预策略建议确实是相关的，其目标是优化这种相互作用。例如，我们在第一年对高危兄弟姐妹的 iBASIS 干预是从旨在改善依恋关系的原始干预改编而来的；但在我们手中，改编有意将二元社会交流作为自闭症的前兆，而不是依恋；同时仍然旨在优化早期开发中的二元社会互动和沟通。在一项纵向随访的 RCT 中，iBASIS 干预成功地将前驱 ASD 症状的严重程度显着持续降低至 3 年（Green 等，2017）。当然，这种二元的改善也可能有益于亲子关系质量以及儿童症状的严重程度，但这与推断依恋不安全感是自闭症的因果前兆相去甚远，我们可以说没有证据。