Early development in autism: How developmental cascades help us understand the emergence of developmental differences

Abstract

Many theories of autism spectrum disorder (ASD) focus on a single system or factor as an explanatory mechanism for autism symptoms and behavior. However, there is growing recognition that ASD is a complex, multisystem neurodevelopmental disorder with origins in prenatal life. Researchers therefore need a conceptual framework that allows examination of the interplay between multiple interacting domains and systems and the ways in which they extend their influence beyond the individual into the surrounding environment. The developmental cascades perspective suggests that even relatively small perturbations in early emerging behaviors in domains that are not traditionally linked may influence subsequent achievements across these areas. In this chapter, we illustrate how a developmental cascades framework can be used to inform the study of atypical development. The developmental cascades perspective provides us with conceptual and methodological tools for considering how variation in children’s real time behavior can provide new insights into sources of variation in their developmental trajectories and outcomes. It also suggests approaches for intervention that leverage targeted skills in novel ways, creating opportunities to support development in other domains and fine-tune caregiver behavior to create powerful moments for infant learning.

Introduction

Autism spectrum disorder (ASD) has historically been conceptualized as a disorder of social communication and interaction (Kanner, 1943). Over the years, a variety of theories have been proposed to explain ASD symptoms and the behavioral profiles exhibited by autistic people¹. For example, some researchers have suggested that a primary driver of behavioral differences observed in autistic individuals is weak central coherence, or a bias towards local rather than global information processing (e.g., Happé, 1999). Others have put forward the notion that differences observed between autistic and neurotypical people are best explained by patterns of functional underconnectivity among brain areas, which results in difficulties integrating information across the neural and cognitive levels (e.g., Just et al., 2007). These theories focus on a single system or factor as an explanatory mechanism for autism symptoms and behavior. This tendency to seek out single mechanisms to explain differences observed in ASD is consistent with traditional approaches in developmental science, in which researchers focus their efforts on characterizing a single behavioral domain and do so in a relatively siloed manner.

However, there is growing recognition that ASD is a complex, multisystem neurodevelopmental disorder with origins in prenatal life. And while it is characterized diagnostically by a wide variety of significant challenges with social communication and interaction (e.g., difficulty sustaining a back-and-forth conversations; challenges establishing and maintaining relationships) and restricted and repetitive behaviors (e.g., repetitive motor behaviors such as hand flapping; an intense interest in a specific subject such as train timetables; American Psychiatric Association, 2013), differences in many other domains of development are also commonly observed (e.g., motor skills, sleep, GI functioning, emotion regulation). Moreover, ASD symptom profiles change across development (e.g., Kim et al., 2018), and new behavioral challenges can also emerge, waxing and waning over time (e.g., conduct, emotional, attention and executive functioning, and mental health symptoms; Stringer et al., 2020).

As this description suggests, autistic people are a highly heterogeneous group, with extensive variability in how core diagnostic behaviors manifest, their severity, and in functioning in other domains. Some autistic people face significant challenges in engaging in social exchanges with others, while others enjoy interacting but have difficulty reading the cues of communicative partners that signal when a conversation should conclude. Some individuals are minimally verbal or entirely nonspeaking, while others may have age-appropriate structural language skills but face difficulties with pragmatics (e.g., turn-taking, prosody, shared conversational topic) and context-appropriate use of language.

This heterogeneity, along with the complexity and multisystemic nature of individual behavior and symptom profiles, suggests that researchers need a conceptual framework that allows for flexible examination of the interplay between multiple interacting domains and systems and the ways in which they extend their influence beyond the individual into the surrounding environment. And because ASD is a neurodevelopmental disability, we need a framework that will allow us to embrace the realities of development and consider not only the effects of developmental time (i.e., across age), but also how developmental change is created in moment-to-moment transactions in real time and how changes in one developmental domain can influence the functioning of others (Thelen, 1992; Thelen & Smith, 1994).

To this end, there has been a recent surge of research highlighting the existence of links between developing systems in infancy and illustrating the downstream, cascading effects of achievements in one domain on infants’ developing skills in other domains (for relevant reviews, see Adolph & Tamis-LeMonda, 2014; Bradshaw et al., 2022; Iverson, 2021). This view has significant implications for how we think about the developmental consequences of early-emerging differences and delays, particularly those that seem quite subtle in nature. The developmental cascades perspective suggests that even relatively small perturbations in early emerging behaviors in domains that are not traditionally linked (e.g., motor, language) may influence subsequent achievements across these areas (Thelen, 2004). In other words, apparently small delays or differences in early motor behaviors may trigger a cascade of direct or indirect effects on the developing communication and language system, such that developmental differences in the downstream domain begin to emerge over time.

In this chapter, we illustrate how a developmental cascades framework can be used to inform the study of atypical development and highlight some of the advantages that it confers for understanding development and developmental process relative to more traditional, siloed approaches. We begin with a brief description of the developmental cascades view and its application to the study of developmental delay and difference. Next, drawing on examples from our own prospective longitudinal studies of infants with an older sibling with ASD, we highlight two principles of the developmental cascades approach—multidirectionality and multiple timescales—that we believe are especially useful for understanding the complexity and variability observed in populations of infants and toddlers with diverse developmental outcomes. We conclude with a discussion of methodological issues in the study of developmental cascades and suggest implications for clinical practice.

Developmental cascades and developmental differences

Development is a highly complex phenomenon. However, developmental scientists often seek to distill this complexity by selecting a particular domain of interest and focusing their work relatively narrowly on describing the nature and shape of developmental change in a particular phenomenon. Although a targeted approach has proven valuable in some ways, yielding rich descriptions of infants’ behaviors, it ignores the fact that in complex systems (like human infants), development does not occur in isolation. At any given moment in time, change is taking place both within and across multiple domains of development as infants engage with an environment that is also constantly changing. Developmental changes in a single domain can have far-reaching, cascading effects on development across domains—even in those that may appear unrelated—and on the environment within which development unfolds. Cascading effects can be direct or indirect; they can be multidirectional; and they can span multiple timescales (moment-to-moment, developmental; Thelen & Smith, 1994). Thus, the notion of developmental cascades provides a framework for conceptualizing the constant interplay between developing infants and their developing interactional environments (e.g., Massand & Karmiloff-Smith, 2015; Masten & Cicchetti, 2010).

The developmental cascades framework also provides a rich perspective from which to study the potential effects of early variability and delay in development. Traditionally, developmental delays and differences have been conceptualized as characteristics of the child (see Karmiloff-Smith, 1998). However, neurodevelopmental disorders such as ASD are inherently complex. Thinking in terms of developmental cascades allows us to consider the potentially far-reaching effects of small, early emerging differences and delays within a domain, across domains, and on the child’s environment.

We made use of the developmental cascades model in our recent studies of infants who have an older sibling with ASD and are therefore at elevated likelihood (EL) for receiving an ASD diagnosis themselves (Ozonoff et al., 2011) and for other developmental delays and differences (Charman et al., 2017). Our general approach has been to enroll infants in longitudinal studies as early in life as possible, and using a dense observation schedule (i.e., biweekly or monthly), follow them until the age of 3 years, when a reliable diagnosis of ASD can be made. At the 3-year visit, we administer a series of clinical and standardized assessments to EL children and assign them to one of three outcome groups: (1) infants who meet diagnostic criteria for ASD (EL-ASD); (2) infants with a history of delayed language development but not ASD (EL-LD); and (3) infants with no clinical diagnoses and who are apparently neurotypically developing (EL-ND). We also include a group of infants who have an older neurotypically developing sibling and no immediate family history of ASD (i.e., typical likelihood infants; TL).

As a group, EL infants exhibit heightened variability in their early developmental trajectories, with many infants (even those who do not go on to an ASD diagnosis) showing delays in the development of foundational motor, communication, and language skills that typically appear during the first year of life (e.g., Jones et al., 2014). Enhanced variation in the onset and production of early emerging abilities has provided us with opportunities to address questions motivated by the developmental cascades framework. Specifically, we have examined how delays in early motor development have cascading effects on the emergence and development of communication and language and on the broader learning environment. In the sections that follow, we describe two principles derived from the developmental cascades framework and review research illustrating their utility for understanding the influence of delayed or differing patterns of development beyond the domain in which they are initially observed.

Principle 1: Multidirectionality, within and beyond the infant

As new motor skills emerge, infants’ relationships with their bodies and their worlds change, and these changes generate a multitude of new opportunities for learning and development. In this section, we describe how new skills have cascading effects both within the infant, by opening doors to higher level skills within and across domains of development, and beyond the infant, by changing interactions with objects, people, and the environment. We discuss the implications of these cascades for infants and young children on the autism spectrum.

Within the infant: Cascading effects in interacting domains

At the most basic level, the development of any new skill is one step on a path towards more advanced skills. For example, infants gain control of the head and neck in the first few months of life, which is foundational for the development of sitting. As infants practice sitting, they expand control of the torso and gain experience balancing in an upright position, skills that lay the groundwork for the later emergence of more advanced locomotor skills (e.g., crawling). This type of within-domain developmental pathway (head control to sitting to crawling) is readily apparent. Less obvious, but equally important, are the ways in which these early motor skills set the stage for a wide range of possibilities for practicing and advancing skills across domains. To date, scholars have used the developmental cascades framework to describe how advances in motor skills influence the development of language, communication, exploration, social interaction, and cognition in neurotypical infancy (e.g., Clearfield et al., 2011; Karasik et al., 2011; Oudgenoeg-Paz et al., 2015; Schneider & Iverson, 2022; Thurman & Corbetta, 2017; Walle & Campos, 2014; West & Iverson, 2022). For infants and young children on the autism spectrum, however, delays or differences in one area of development may alter opportunities for subsequent development both within and across domains.

The emergence of independent sitting, for example, is a transformational moment in the first year of life, bringing about a host of changes in infants’ experiences with objects and people by providing new opportunities for exploration and interaction. Research on neurotypically developing infants has detailed several key changes in infant behavior that emerge in tandem with the acquisition of independent sitting. As infants transition from lying (supine or prone) to the upright position, their field of view expands to a panorama of their surroundings and the objects and people within them (Bertenthal & von Hofsten, 1998; Luo & Franchak, 2020). This expanded view (along with the freeing up of hands) provides new opportunities for infants to explore objects in more sophisticated ways by coordinating looking with other manual actions (e.g., Soska et al., 2014) and engaging in joint looking to and manipulation of objects with caregivers (e.g., Franchak et al., 2018; Kretch et al., 2021). Sitting also repositions the vocal apparatus and respiratory system in ways that support production of reduplicated babble (i.e., vocalizations containing repeated syllables; e.g., [bababa]; Yingling, 1981). Thus, from a developmental cascades perspective, the development of a single new motor skill—sitting—triggers a series of advances in multiple domains of development.

As a group, EL infants begin to sit independently later than their TL peers (e.g., Jarvis et al., 2020; Nickel et el., 2013). And even once infants attain the skill, they spend less time in the sitting posture compared to TL infants (Leezenbaum & Iverson, 2019), suggesting that for EL infants, consolidating the skills required for stable independent sitting may occur over a more protracted period of time. Additional evidence for this possibility comes from the observation that EL infants spend less time grasping objects when sitting independently than when sitting with support, whereas TL infants spend similar proportions of time grasping in these postures (Mlincek et al., 2022). Indeed, EL infants may be particularly impacted by the demands of balancing and holding the trunk upright while also grasping objects; a type of multitasking that is inherently challenging and requires practice for successful execution (see Berger et al., 2017).

For infants with motor delays, the extended time period required for sitting consolidation may limit the cascading effects of sitting achievement on object manipulation (Mlineck et al., 2022), with the result that EL infants may take longer to exhibit the advances in object actions that generally occur with the transition to independent sitting. Challenges in object exploration and manipulation may, in turn, impact opportunities for experiences that influence the development of communication, cognition, language, and play (Babik et al. 2022; Zuccarini et al., 2017). Thus, what may initially present as a small delay in a single developmental domain (in this case, the motor system) can have cascading effects that extend well beyond that domain.

The example of cascades from ASD to sitting to other domains illustrates cascading developmental effects that occur within the infant. However, these changes do not take place in isolation. We now turn to a discussion of how early delays and differences in development may extend beyond the infant, influencing both infant-caregiver interactions and the broader relationship between infants and their immediate interactional environment.

Beyond the infant: Cascading effects in infant-caregiver interactions

Advances or delays in development impact infants’ opportunities to receive input from the environment, as well as the nature of the input. Infants are born into a world surrounded by language, and their daily activities are infused with language produced by caregivers. For many years, when researchers talked about the importance of caregivers in infant development, the focus fell largely on how caregiver language input impacts infant development (e.g., Topping et al., 2013). Indeed, an abundance of research shows that caregiver input to infants and children on the autism spectrum is positively related to language development (for reviews, see Bottema-Beutel & Kim, 2021; Swanson, 2020). For example, caregiver MLU, richer home language environments, and more conversational turns are positively associated with child language development (e.g., Bang & Nadig, 2015; Choi et al., 2020; Fusaroli et al., 2019; Swanson et al., 2019), and translations of child gestures facilitate word learning (Dimitrova et al., 2016). It is clear from this research that autistic children, like their neurotypical peers, benefit from the parental input they receive.

More recent work, however, has highlighted the central and active role that infants play in shaping caregiver communication and behavior (e.g., Kuchirko et al., 2018; Tamis-LeMonda et al., 2018). Research suggests that neurotypically developing infants are armed with attention and learning mechanisms that allow them to reap the benefits of caregiver language input. As infants acquire new skills, caregivers adapt their communicative behaviors and respond in ways that encourage production of more advanced skills in their infants.

From a developmental cascades perspective, this bidirectional set of interactions between caregiver and child is not a simple loop (e.g., child language prompts caregiver response which prompts child language). Instead, it can be thought of as an ever-widening ripple effect in which caregiver input changes and expands as the child’s behaviors advance, and in turn, the child’s behavioral repertoire advances and expands in response to the caregiver’s more complex input. This framework provides an approach for understanding how early developmental delays and differences in infants and young children on the autism spectrum may have far reaching consequences that impact trajectories of development across multiple domains.

Several studies have reported that relative to caregivers of TL infants, caregivers of EL infants² are similarly responsive to their infants’ social and communication behaviors. However, early delays and differences in EL infants (particularly EL-ASD infants) across the motor, social, and communication domains mean that they provide their caregivers with fewer opportunities to respond (e.g., Calabretta et al., under review; Choi et al., 2020; Leezenbaum et al., 2014; Warlaumont et al., 2014; West, 2019). This dynamic has been demonstrated repeatedly in both the communication and the motor domains.

For example, caregivers are more likely to provide “translations” (i.e., labeling the referent of their child’s gesture) to infant pointing and showing gestures than to early emerging gestures like reaching or giving (Leezenbaum et al, 2014), and they are also more likely to respond to speech-like vocalizations (e.g., babbles, coos) than non-speech (e.g., laughing, crying, burping, coughing) vocalizations (Warlaumont et al., 2014; Warlaumont, 2020). As a result, EL and EL-ASD infants, who tend to show delays and differences in the production of these more advanced communicative behaviors, may miss opportunities to receive rich, contingent responses from caregivers (Leezenbaum et al., 2014; Warlaumont et al., 2014).

Caregiver responses to infants’ gestures and vocalizations, in turn, impact infant communication, both in the moment and over the long term (e.g., Elmlinger et al., 2019). When caregivers respond to their infants’ speech-like vocalizations, infants are more likely to produce a subsequent speech-like vocalization (Warlaumont et al., 2014). Adult translations of infant gestures facilitate word learning in both neurotypical and autistic children (Dimitrova et al., 2016). In this way, differences and delays in early communication skills impact opportunities to receive parental input that could enhance and expand those skills and have cascading effects on future development.

Caregiver input is also shaped by infant behaviors outside the realm of communication. Research on neurotypical development suggests that advances in infants’ motor skills (e.g., the onset of independent sitting or walking) impact caregiver behavior and the input infants receive from caregivers in ways that influence opportunities for development. For example, we observed that independent sitting shapes how infant-caregiver dyads co-construct the physical spaces of object play (Schneider et al., 2022). Whereas mothers predominantly structured the physical organization of interactions at younger ages (e.g., by facing their infants from above while infants were supine), infants increasingly shaped dyadic positioning (i.e., how mothers’ and infants’ bodies were positioned relative to one another) as they got older, and especially after learning to sit independently. This resulted in more time spent in a broader body configuration, specifically at right angles. A larger interaction area (a contribution of infant sitting) has immediate consequences for learning. Specifically, caregivers are significantly more likely to provide infants with rich cognitive learning opportunities when infants are sitting independently compared to when they are lying supine (Kretch et al., 2021).

Importantly, this phenomenon is not unique to sitting. Caregivers are also more likely to provide contingent language and gestures when infants walk compared to when they crawl (Schneider & Iverson, 2022) and are more likely to respond to infant communicative bids that are paired with locomotion (i.e., moving bids) than to bids from a stationary position (e.g., Karasik et al., 2014; Toyama, 2020). Research has also shown that caregivers had significantly more opportunities to respond to moving bids after their infants started walking, suggesting a tight link between patterns of infant action and subsequent caregiver communication (West & Iverson, 2021).

Infants later diagnosed with ASD are delayed in the onset of walking (see West, 2019 for a meta-analysis), and a critical implication of this delay is that the input they receive may differ from that of similarly aged peers with more advanced motor skills. Thus, for example, in an investigation of how 18-month-old EL infants used walking to initiate social interaction (i.e., to approach their caregivers, carry an object while approaching, or share an object by producing a moving bid), we discovered that EL-ASD infants walked less frequently, and by extension, used walking for purposes of social interaction less often than EL infants without ASD and TL infants (Calabretta et al., under review). And although caregivers of all infants were similarly and highly likely to provide a verbal response when infants produced moving bids (i.e., approaching to share an object with a gesture), only caregivers of EL-ASD infants were significantly more likely to respond to their infants’ approaching behaviors. Importantly, however, EL-ASD infants produced all social actions less often than their peers, with the net result that they provided caregivers with fewer opportunities to respond.

Evidence highlights the nuanced, multidirectional nature of the relations between children’s developing language skills and caregiver input. Relative to neurotypical peers, toddlers on the autism spectrum tend to spend less time in coordinated joint engagement (i.e., playing together with the same object as a caregiver and making eye contact during the interaction; Roemer et al., 2022). However, their caregivers increased the rate of object labels provided in this engagement state from 12 to 18 months, whereas caregivers of EL infants without ASD and TL infants did not. It may be that caregivers of EL-ASD infants are sensitive to emerging delays in their child’s communication and language abilities and adapt their speech accordingly, providing more labels during infrequent but salient moments of coordinated joint engagement in an effort to address these delays.

Other studies provide further evidence for the idea that caregivers of EL children, particularly those later diagnosed with ASD, alter their input to their children in ways that may support child attention and learning. For example, EL caregivers are more likely to use multimodal input than TL caregivers, pairing speech with physical contact (Kadlaskar et al., 2020), and are more likely to position themselves to be in closer proximity to their infants during play (Srinivasan & Bhat, 2020). Moreover, mothers of children with ASD provide more gestures during word learning tasks (Yoshida et al., 2019) and use more repetitions of labels when teaching novel object names than parents of neurotypical peers (Bani-Hani et al., 2013). But less clear is whether and how autistic children learn from this input. For example, when children are just beginning to actively coordinate eye contact and joint play with objects, they may not be ready to process and learn from abundant, salient labels in the same way as their neurotypical peers (e.g., Roemer et al., 2022).

In summary, delays or differences in early emerging skills (e.g., communication, movement, social initiation) can have far-reaching effects across domains of development and beyond the infant. These differences can cascade into diverging opportunities for interaction and communication, alterations in the language learning environment, and variations in interactions with the physical environment. As opportunities diverge, so does development, and thus what began as perhaps a subtle difference in one domain can cascade across the developing system in multiple directions. Importantly, the complexity of the cascades framework requires researchers to consider how development occurs on multiple timescales—within moment-to-moment interactions between the infant and the environment and across weeks, months, and years as those interactions accumulate. In the next section, we discuss why studying development on multiple timescales is essential for revealing cascading effects.

Principle 2: Timescales, from moment to moment and across development

The developmental cascades framework appeals to researchers because it offers an approach for considering the question of why new skills often emerge concurrently in developmental time. Infants gain proficiency in many foundational skills, and their progress across domains often coincides. For instance, when infants begin to walk, their word learning often quickens (e.g., He et al., 2016; Walle & Campos, 2014; West et al., 2019). When infants begin to manipulate objects and move, their spatial understanding skills improve (e.g., Dosso & Boudreau, 2017; Oudgenoeg-Paz et al., 2016); Soska et al., 2010). And when infants’ sleep schedule matures, their memory consolidation changes (e.g., Tham et al., 2017). Indeed, a large body of research—including studies of neurotypical and autistic children—finds correlations in the timing of skills across seemingly separate domains. One possibility is that progress in these domains is functionally related. However, other explanations could account for simultaneous growth across domains (for example, general maturation may drive change across a suite of domains, including motor, cognitive, and language; e.g., Zelazo, 1983).

Correlations among skill acquisitions do not address the question of mechanism. How, in the course of infants’ everyday experiences, does the expression of a behavior in one domain relate to learning in another? What is it about walking upright that influences infants’ word learning? How does manipulating objects in the moment support spatial cognition? To understand whether and how a developmental cascade unfolds, researchers must observe behavior on a moment-to-moment basis, in the context of infants’ natural environments and routines, and piece together how multiple domains interact in real time and in real life. Understanding clearly the “why” behind developmental cascades is especially consequential for researchers studying autism, because findings may be translated into clinical practice and intervention.

Trends over developmental time are explained in real time

When researchers observe a connection between two skills over developmental time, the next step is to uncover the real time behaviors that underlie the association. Consider the development of walking and talking. Common sense suggests that learning to walk and talk are very different feats and therefore, success in one skill is likely unrelated to success in the other. However, studies have documented an unexpected pattern in the emergence of walking and word learning over developmental time. Among neurotypical infants, the pace of word learning increases precisely when infants begin to walk, regardless of the age at which walking onset occurs (He et al., 2015; Walle & Campos, 2014; West et al., 2019). In one such study, we asked caregivers to report the words their infant understood each month over a 7-month period, during which time infants transitioned from crawling to walking (West et al., 2019). Before learning to walk, neurotypical infants’ receptive vocabularies grew at a rate of about 11 new words per month. But after they began to walk, the pace of word learning more than doubled, increasing to about 28 new words each month. However, for EL-ASD infants, the rate of word learning did not significantly change during the transition from crawling to walking and instead remained relatively flat.

Why did walking benefit word learning for neurotypical infants but not EL-ASD infants? In a series of follow-up studies, we leveraged naturalistic observations of infants playing at home with their caregivers across the transition from crawling to novice walking and documented dyads’ moment-to-moment locomotor and communicative behaviors. Overall, EL-ASD infants crawled and walked just as frequently as other EL and TL infants, and they also retrieved and carried objects just as often (West, 2019). But among TL infants, the frequency of spontaneously produced gestures increased substantially during walking months compared to the prior crawling months (West & Iverson, 2021). That is, after they began walking, TL infants increasingly pointed, showed toys, reached to request desired objects, or handed objects to social partners more frequently than they did when they were pre-walkers.

In addition, TL infants often paired their communication with locomotion in the moment—approaching their caregiver to vocalize or gesture (e.g., walking up to a caregiver and handing over a toy; West & Iverson, 2021). By contrast, EL-ASD infants gestured infrequently across the entire 7-month period, and growth in their spontaneous gesture production did not change with the emergence of walking. Indeed, infants with ASD very rarely coordinated locomotion with socially directed vocalizations and gestures (West, 2019, West et al., under review).

Caregivers of all infants frequently responded with language input when their infant paired locomotion with a communicative behavior (e.g., “is that your cup?” as infant approaches to show a sippy cup), and they responded less often when their infant communicated from a stationary position (a behavior which may more easily be overlooked; e.g., Karasik et al., 2011). Because neurotypical infants produced far more moving bids after beginning to walk, walk onset marked a substantial increase in these bid-and-response exchanges between infant and caregiver. In contrast, EL infants with ASD produced far fewer moving bids than neurotypical infants, so their caregivers had fewer opportunities to respond in kind. As a result, these infants experienced substantially fewer bid-and-response exchanges with caregivers than did neurotypical infants. The differing rates of bid-and-response exchanges likely accounts, at least in part, for why neurotypical infants experienced a boost in vocabulary when they began to walk, while word learning remained stable for infants with ASD. Thus, caregivers’ contingent responses offer rich opportunities for word learning, particularly when the language input is contextually connected to the infant’s moment-to-moment behavior (e.g., Custode & Tamis-LeMonda, 2020; Schneider & Iverson, 2022; Suanda et al., 2019; Suarez-Rivera et al., 2022; West et al., 2022; Yu & Smith, 2012).

The connection between walking and talking offers an example of how behavioral cascades in the moment can shed light on broader trends over long stretches of developmental time. Infants with and without ASD showed very distinct patterns of month-by-month vocabulary growth as they transitioned from crawling to walking. And trends in vocabulary growth were closely mirrored in infants’ minute-by-minute social interactions. Here, infants’ in-the-moment walking steps had immediate consequences for their social interactions—walking provided autonomy for infants to approach caregivers and thereby elicit rich language input, which infants with vs. without ASD produced at very different frequencies. Even small fluctuations in these exchanges can accumulate over time to influence infant learning.

Methods for studying developmental cascades

A developmental cascades approach buys much in the way of asking broader questions that extend beyond the study of single behaviors in siloed domains. This is particularly important for researchers interested in ASD because a primary goal is not only to examine the emergence of ASD, but also to identify and understand sources of the extensive variability observed in the developmental trajectories of autistic children. Studying behavior through the lens of developmental cascades requires methodological innovation and flexibility so that researchers can capture behavior at multiple levels and on multiple timescales. Here we provide several methodological suggestions for researchers to consider when taking a developmental cascades approach to their research questions.

First, as previously discussed, the developmental cascades framework requires sampling behavior on multiple timescales and at multiple levels of analysis. In our work, we have leveraged both chronological age-based designs (e.g., Leezenbaum & Iverson, 2019; Roemer et al., 2022; Schneider et al., 2022) and milestone-based designs, in which observations are anchored to the onset of a particular skill without regard to infant age; Jarvis et al., 2020; Schneider & Iverson, 2022; West et al., 2019). Moreover, the issue of timescale goes beyond the selection of ages and timepoints. Developmental cascades hinge on the concept that larger changes over months and years arise from smaller changes over the course of days, hours, and even minutes and seconds. Examining the exchanges that comprise moment-to-moment interactions and ways in which they change over time are necessary for building the larger picture of development (see de Barbaro, 2019).

Second, researchers must use a variety of tools to observe and gather data on infant behavior and developmental change. Traditionally, video has been a cornerstone of observational science (see Adolph et al., 2017). Video captures observable behaviors and allows researchers to identify and classify them and compile data about the unfolding of different actions and how they change over time. However, combining video observation with other technologies that permit collection of information at multiple levels of analysis provides more fine-grained data on aspects of behavior that shift rapidly (e.g., eye gaze and looking behavior) or are unobservable to the human eye (e.g., kinematics of limb movement and postural control).

As researchers have noted, the immense heterogeneity in the development of EL infants makes the detection of differences and delays from observable behavior alone a difficult task. It is sometimes the case that global behavioral differences are quite apparent. But it is more often the case that groups of infants are better distinguished by differences in very subtle features of behaviors (e.g., postural sway, trajectories of arm movements, the distance between successive walking steps) than by easily observable features. Leveraging new technologies gives researchers additional tools with which to track how behaviors are organized in time, and how that organization may vary in infants and toddlers with autism.

Finally, standardized assessments have been widely used in research on EL infants (e.g., Mullen Scales of Early Learning). These tools identify infants’ abilities in core areas of development and compare them to age-based norms from samples of neurotypically developing children. However, standardized assessments are designed to capture what infants are capable of doing under ideal conditions—typically during highly structured tasks (e.g., experimenter gives infant a tightly sealed jar with a desirable toy inside, in an attempt to elicit a requesting bid from the infant)— rather than capturing how infants behave spontaneously in naturalistic contexts. For example, although two infants may both request help from an experimenter to open a jar, the frequency and content of their spontaneous requesting behaviors may vary substantially in naturalistic settings. That is, the expression of a certain behavior in a standard assessment tells us only that an infant can perform the skill; but it tells us virtually nothing about how often infants produce it in daily life or the contexts in which it is used (see Tamis-LeMonda et al., 2017, for a similar argument).

Standardized assessments are not always sensitive to the wide spectrum of language ability in autistic children. Natural language sampling measures are more sensitive to change for minimally verbal children (see Barokova & Tager-Flusberg, 2020 for a review). And structured settings may not fully capture children’s strengths. Indeed, Kover et al. (2014) found that language samples collected during the ADOS yielded fewer words, shorter utterances, and fewer requests, comments, and instances of turn-taking than language samples collected from unstructured play. Thus, researchers interested in examining cascades should observe behaviors that occur in the complexity of everyday interactions at home and to do so frequently in order to capture change over time as it is unfolding.

Implications for clinical practice

Interventions for autism have largely focused on the core features that differentiate autistic from neurotypical children. Historically, this involved intensive behavioral interventions focused on increasing discrete skills such as eye contact and gesture use and decreasing behaviors such as hand flapping and other repetitive behaviors. Interventions for domains that autistic individuals also struggle with but are not considered core features (e.g., motor skills) were relegated to the host of early intervention services that children received from different providers—including physical therapy, occupational therapy, and speech therapy. Ideally, providers would coordinate treatment and treatment goals with one another to form a cohesive intervention team, but such truly integrated care is difficult to implement and rarely achieved. More recently, researchers have called for such integration, for example calling for motor skills to be incorporated throughout screening, evaluation, and treatment planning for ASD (see Zampella et al., 2021 for review).

These earlier intervention approaches also generally did not attend to the contexts in which targeted behaviors occurred. Today, a more common practice is the use of naturalistic developmental behavioral interventions (NDBIs), which place the child in context and use the child’s interests to intervene and develop skills. The Early Start Denver Model (ESDM) is one well-known example of an NDBI, and randomized controlled trials show such interventions to be effective (e.g., Dawson et al., 2010; Waddington et al., 2016). There is currently a wide range of NDBIs, each with their own focus and techniques (see Bruinsma et al., 2020 for a review), but all are founded on using behavioral principles and embedding these behavioral learning opportunities in the context of natural, developmentally appropriate contexts—for example, during everyday play. NDBIs take a developmental approach to behavior change and are thus a natural target for integrating a developmental cascades perspective into autism interventions. Below we suggest implications for clinical intervention based on the principles of multidirectionality and multiple timescales described above.

First, multiple domains of development must be considered in tandem. There is a need for greater awareness among intervention providers that subtle delays in one area can have downstream influences on the development of skills in other domains. Recall that EL infants sit independently approximately two weeks later than their TL peers; and when they do sit, they spend less time grasping objects when positioned in independent sitting. Thus, if there is a delay in one domain, clinicians must consider the multiple streams of affected behavior both within and beyond the infant, and where possible, find alternative methods for providing these opportunities for learning.

In this case, a central goal would be to boost independent sitting, a typical area of focus in physical therapy. However, as infants are building and consolidating skills necessary to maintain independent sitting, they may be missing out on salient opportunities for learning from objects, as they spend less time grasping and exploring objects. Thus, an equally important companion goal may be to discover ways to create opportunities for object exploration. For example, if an infant successfully grasps objects when sitting with support, providing targeted time in supported sitting may facilitate object exploration and support the cognitive opportunities that come along with it.

Second, behavior does not exist in isolation. It is critical to consider developing infants in the context of their environments and their multidirectional influences on learning and opportunities for learning. In the example above, when providing targeted time in supported sitting to facilitate object exploration, coaching caregivers to label an object when the infant is holding it may generate the rich language input that is typically directed to infants as they hold, look at, and explore objects (e.g., West & Iverson, 2017). NDBIs have made great strides in incorporating caregivers in the intervention process and facilitating learning in the context of naturalistic, everyday interactions in the home. Such models have also been implemented as parent-mediated interventions using the same techniques and coaching parents to scaffold and create behavioral opportunities at every possible moment, day-to-day, in their children’s everyday routines and activities (Kasari et al., 2015; Rogers et al., 2012). This approach allows learning opportunities to extend well beyond the limited hours of intervention and occur whenever the developing child interacts with caregivers.

The literature reviewed above also underscores the fact that behavior in the context of the caregiver-child dyad is multidirectional and ever-changing. That is, change is dramatic and rapid in the first few years of life, and as infants’ skills change, caregivers must modify their behaviors to accommodate their infants’ new behaviors and abilities. In real life interactions, caregivers respond to the infant in front of them, adapting their behavior to their child’s perceived abilities. For caregivers of very young children with delayed communicative and language development, this adaptation may involve capitalizing on any attempts at communication, even if they are less developmentally advanced (e.g., Calabretta et al., under review). And when infrequently produced behaviors do occur, they may present caregivers with a rare opportunity to engage with their child in a new way. Caregivers may respond by filling these moments with language input, perhaps in part, to address the child’s delays (Roemer et al., 2022).

It remains unclear whether moments of intensive input are beneficial for learning for young children with ASD. The assumption (based on studies of neurotypical development) is that more input from caregivers is better for learning—an assertion supported by findings showing associations between high rates of contingent input from caregivers and better language scores in infants. But it may be that in some contexts and/or for some children, simply providing more input is not optimal. By studying patterns of such behaviors at multiple timescales to uncover the cascade of moment-to-moment and multidirectional influences between children and their environments, we can better understand the types of input that are potentially most beneficial and the contexts within which they are most effective.

Finally, and perhaps most importantly, existing interventions are couched in the assumption that neurotypical behavior is the desired outcome for autistic people—and thus goals are often focused on increasing eye contact, building typical communication (i.e., speech), and reducing stereotyped and repetitive behaviors. But this assumption can be dangerous and have unintended consequences. For example, for some autistic individuals, avoiding eye contact might be a way to reduce cognitive load, stereotyped behaviors may be a self-regulatory tool, and echolalia may serve a communicative function (see Jaswal & Akhtar, 2019 for further discussion of these ideas). The developmental cascades perspective embraces the variability and flexibility inherent in developmental process and suggests that there are multiple paths to a given behavioral state or outcome.

Using eye contact as an example, a long-standing assumption from studies documenting associations between joint attention and child language skills (e.g., Charman et al., 2003; Mundy et al., 1990) is that eye contact and gaze following are necessary skills to share attention to an object (and subsequently, to benefit from language input about the object of shared attention). But research using novel methods (e.g., head-mounted eye-tracking) and studying the moment-to-moment dynamics of interaction shows that there are many routes to achieving shared attention to an object. Neurotypical infants use a variety of other routes into shared attention, such as using caregivers’ hands as a cue to locate the caregiver’s focus of attention (e.g., Yu & Smith, 2013). And autistic children use similar routes into shared attention that do not require mutual eye contact or following the caregiver’s gaze (Yurkovic-Harding et al., 2022). Thus, researchers and clinicians must consider what the desired outcome is for a given intervention. If the desired outcome is to increase language and communication, there may be multiple routes to that outcome, not all of which require increased eye contact. By considering the possibility of alternate routes to an outcome like language, clinicians can capitalize on the flexibility inherent in development and build on the child’s strengths.

Conclusion

In 1998, Annette Karmiloff-Smith published a paper entitled Development itself is the key to understanding developmental disorders. Nearly 25 years later, this statement stands as a powerful reminder to scientists that development is not merely a simple march forward in time, but a highly complex process. It involves multidirectional interactions between multiple systems, each of which involves multiple subsystems. And these subsystems, systems, the environment, and the interactions between them are constantly changing over time.

In this chapter, we have illustrated how even small, early appearing differences in one system can have downstream cascading effects on the development and functioning of interconnected systems, on caregivers, and on the environment. Understanding development in neurodevelopmental disorders requires embracing this complexity and adopting a theoretical framework that incorporates its influence at multiple levels of analysis and on multiple timescales. The developmental cascades perspective provides us with conceptual and methodological tools for evaluating the enhanced variability that characterizes atypical development. It allows us to consider how variation in children’s real time behavior can provide new insights into sources of variation in their developmental trajectories and outcomes. And it suggests approaches for intervention that leverage targeted skills in novel ways, creating opportunities to support development in other domains and fine-tune caregiver behavior to create powerful moments for infant learning.