Communication changes when infants begin to walk

Abstract

Learning to walk allows infants to travel faster and farther and explore more of their environments (e.g., Adolph & Tamis-LeMonda, 2012). In turn, walking may have a cascading effect on infants’ communication and subsequent responses from caregivers. We tested for an inflection point—a dramatic shift in the developmental progression—in infant communication and caregiver responses when infants started walking. We followed 25 infants longitudinally over 7 months surrounding the onset of walking (Mean walk onset age = 11.76 months, SD = 1.56). After learning to walk, the pace of gesture growth (but not vocalization growth) increased substantially, and infants increasingly coordinated gestures and vocalizations with locomotion (e.g., by walking to a caregiver and showing off a toy bear). Consequently, caregivers had more opportunities to respond contingently to their infants during walking months compared to crawling months (e.g., “What did you find? Is that your bear?”). Changes in communication were amplified for infants who began walking at older ages, compared to younger walkers. Findings suggest that learning to walk marks a point in development when infants actively communicate in new ways, and consequently elicit rich verbal input from caregivers.

New motor skills create new opportunities for infants to interact with the environment (e.g., Adolph & Tamis-LeMonda, 2014; Campos 2000; Iverson, 2010; Needham & Libertus, 2011). For instance, learning to walk allows infants to travel greater distances at faster speeds compared to crawling (Adolph et al, 2012). This newfound efficiency allows walkers to access distant destinations, objects, and people (Karasik, Tamis-LeMonda & Adolph, 2011; Adolph & Tamis-LeMonda, 2014; Biringen, Emde, Campos & Appelbaum, 2008; Mahler et al., 1975). Walking may also have a cascading effect on communication (e.g., Iverson, 2010). Cross-sectional studies show that walkers approach, bid for caregivers’ attention, and gesture more often than same-aged crawlers (e.g., Clearfield, 2011; Karasik, Tamis-LeMonda & Adolph, 2011; Toyama, 2018). If indeed walking alters infant communication, then walk onset should mark an inflection point—that is, a meaningful shift in the developmental progression—in communicative development. Here, we prospectively documented infants’ spontaneous gestures and vocalizations over seven months during the transition from crawling to walking. We also tested for corresponding changes in caregivers’ responses to infant communication. Our prospective longitudinal design allowed us to compare growth before and after walk onset, and accordingly, to detect whether an inflection in communication occurred with the advent of walking.

Walking and communication

Infants transition from crawling to walking after their first birthday (e.g., Adolph & Berger, 2006). The incentives to do so are clear. Walking allows infants to travel up to 700 meters per hour, much faster and farther then they previously locomoted as crawlers (Adolph & Tamis-LeMonda, 2012). Research indicates that practice walking in and of itself is rewarding to infants, and provides opportunities for exploration (Hoch, O’Grady & Adolph, 2019; Hoch, Rachwani & Adolph, 2020). Rather than setting out for specific destinations, infants largely walk to forage in their environment, discovering interesting objects and social partners along the way. In addition, walkers gain an elevated vantage point by standing upright. Whereas crawlers’ view is dominated by the floor right in front of them, walkers have a broad view of the room as they move (Kretch, Franchak & Adolph, 2014). Thus, walking enhances infants’ ability to go, see, and explore their environment (see Adolph & Tamis-LeMonda, 2014, for review).

Recently, researchers have investigated whether walking has a cascading effect on language growth, with mixed results. Several studies indicate that word learning accelerates after infants begin to walk (He, Walle & Campos, 2016; Walle & Campos, 2014; West, Leezenbaum, Northrup & Iverson, 2018). However, others did not replicate this phenomenon (Moore et al., 2019). It is worth noting that when walking emerges—at around 11 to 13 months—infants produce very few words. Most 13-month-old infants are just beginning to say their first words (Carey, 1978; Fenson, 1994; Frank, 2016). More often at this age, infants express themselves through gestures and non-word vocalizations (e.g., Bates, 1976; Bates, Benigni, Bretherton, Camaioni, & Volterra, 1979; Oller et al. 1999). Thus, understanding the connection between walking and communication requires information on how prelinguistic communication changes at walk onset.

Prior work indicates that walking infants gesture, vocalize, and look to their caregivers more often than crawlers do (Clearfield et al., 2008; Clearfield, 2011; Walle, 2016; Yamamoto, Sato & Itakura, 2020). For example, Clearfield (2011) reported that socially-directed vocalizations and gestures doubled when infants were walking, compared to a prior session in which infants crawled or moved in a baby-walker. Why would walking influence the frequency of preverbal communication? Perhaps the ability to approach caregivers with ease enables walkers to initiate social interactions more often than crawlers, for whom locomotion is less efficient. Indeed, walkers frequently locomote back-and-forth between caregivers and other destinations (Biringen, Emde, Campos & Appelbaum, 2008; Mahler et al., 1975). Moreover, walkers often pair their social bids with locomotion (e.g., by walking up to a caregiver and showing a toy), whereas crawlers tend to bid from a stationary position (e.g., by gesturing in place; Karasik, Tamis-LeMonda & Adolph, 2011; Toyama, 2018).

It is possible that communication and walking skills simply undergo parallel advances that are both driven by general developmental change (i.e., infants improve in both domains over time, but the improvements are unrelated). Alternatively, walking and communication may be functionally related via a developmental cascade, in which walking creates new opportunities for infants to communicate in new ways. If walking instigates such a cascade, then the onset of walking should mark an inflection point—a point when growth substantially increases—in the progression of communicative development. To date, no studies have tested whether the pace of communication growth increases after walk onset, over and above the prior month-to-month growth.

Walking and caregiver responses

Changes in infant communication likely influence caregivers’ responses to their infants. Caregivers are highly responsive to infant behaviors—up to 50–70% of caregiver utterances are responses to infant actions (e.g., Bornstein, Tamis-LeMonda, Hahn & Haynes, 2008; West & Rheingold, 1978). Infants’ gestures and vocalizations elicit rich verbal responses from caregivers (e.g., “is that your bear?” as an infant shows off a toy bear; Leezenbaum, Campbell, Butler & Iverson, 2014; Paavola, Kunnari, Moilanen, & Lehtihalmes, 2005; Tamis-LeMonda et al., 2001). Caregivers are especially responsive to infants’ “moving” communications; i.e., behaviors that are temporally paired with locomotion (Karasik, Adolph & Tamis-LeMonda, 2014; Toyama, 2018). Moving communications may be more salient to caregivers and reflect a high level of directedness on the part of the infant. If indeed infant communications change after walk onset, becoming more frequent and incorporating movement, then caregiver responses likely also change. No study to our knowledge has tracked caregiver responsivity longitudinally during the transition to walking.

Current study: Is there a shift in infant and caregiver communication at walk onset?

Prior literature suggests that walking has a cascading effect on infant communication, influencing how often infants gesture, vocalize, and approach caregivers, and in turn, caregiver responses (e.g., Clearfield et al., 2008; Clearfield, 2011; Walle, 2016; Yamamoto, Sato & Itakura, 2020; Karasik, Tamis-LeMonda & Adolph, 2014). However, there are important limitations to this work. First, many studies used cross-sectional designs, comparing age-matched samples of crawlers and walkers. This approach is confounded by infants’ general developmental ability. Precocious walkers may also just happen to be precocious communicators. Second, most longitudinal studies on this topic have relatively limited observation schedules (i.e., two or three observations), and therefore provide no baseline for the rate of change prior to walking. Absent a baseline measurement, it is impossible to determine whether a change in communication occurred with walking per se, or whether there was simply a continuation of communicative development over time. Third, previous longitudinal research has considered time relative to infant chronological age (e.g., at 10-, 12-, and 14-months old), meaning that infants become “walkers” at different points in the observation schedule. A prospective milestone-based approach, where observations are aligned by time relative to the onset of walking (e.g., two months before walk onset, one month before walk onset, and so on) would clarify how development changes in relation to walk onset.

Finally, it is unknown whether the timing of walk onset influences communication. Does walking have a greater impact on communication for younger or older infants? It is possible that younger infants—less mature communicators—have more room to grow in this domain, and consequently stand to benefit more from walking than older infants. Alternatively, older infants—more mature communicators—may be better situated to take advantage of the opportunities for communication that walking creates.

To address these open questions, we followed 25 infants longitudinally for seven months as they transitioned from crawling to walking. We observed infants’ spontaneous communication during naturalistic at-home interactions with their primary caregiver. In particular, we measured infants’ gestures and vocalizations, the extent to which gestures and vocalizations were paired with locomotion, and the contingent verbal responses they elicited from caregivers. The milestone-based longitudinal design allowed us to determine whether growth in these behaviors shifted when infants began to walk, and whether infants’ age at walk onset impacted their trajectories.

Methods

Participants

We collected data from 25 infants (15 male) born from full-term, uncomplicated pregnancies and from monolingual English-speaking households. Nine infants were first-born, and 16 had at least one older sibling. Twenty-four infants were Caucasian and one was Asian. Caregivers were highly educated—the majority held a college degree or completed some college. Mothers were M = 31.92 (SD = 4.85) and fathers were M = 33.08 years old (SD = 4.00).

Procedure

As part of a larger study, infants were followed longitudinally from 2 to 19 months of age. This study focused on a window of seven monthly visits surrounding infants’ walk onset (regardless of their chronological age). The window began with a visit occurring 4 months prior to walk onset and ended with a visit occurring 3 months after walk onset. Thus, seven timepoints were included per infant, and the mid-point was the visit just prior to walk onset.

The onset of walking was established for each infant via caregiver report. Upon enrollment in the study, caregivers were given a calendar to track their infants’ achievement of major early motor milestones, including walking. Walk onset was defined as three consecutive, alternating, and independent steps with no support from furniture or a caregiver. At every session, the experimenter would ask whether the infant had attained walk onset. Once reported, the experimenter confirmed that the baby had indeed met the above criteria. Note that although walk onset could have occurred between infants’ monthly visits, we will refer to the first month in which infants met this criterion as their designated “walk onset”. All infants had started walking before their final 19-month visit.

During visits, infants were videotaped in their homes for approximately 45 minutes of structured and unstructured play activities. Generally, participant’s homes were medium to large suburban houses. Visits typically occurred in participants’ living rooms with a wide selection of toys nearby. Two or three researchers were present during visits. One researcher video-recorded the entire session, taking care to not interact with the infant. Another researcher administered paperwork, implemented study protocols, and timed the research segments. In addition, when necessary, a third researcher attended visits to occupy older siblings and played with them in a different room, away from the infant-caregiver dyad. We selected the first 10-minute segment of each visit to be coded, during which caregivers were instructed to play with their infant as they typically would. For 166 of the 168 sessions, the 10-minute segment was continuous with no pauses. However, for 2 visits the segment was broken up by a pause (i.e.., two segments lasting 5 minutes) because the child’s movements were constrained (e.g., in a high-chair, play pen or infant seat).

Coding & Data Reduction

Coding was completed in version 4.8 of Elan for Windows (http://tla.mpi.nl/tools/tla-tools/elan; Lausberg & Sloetjes, 2009). Coding was completed by one primary and five secondary coders. Secondary coders were trained by double-coding videos with the primary coder until they met threshold scores on all variables for 3 consecutive videos (mean percent agreement ≥ 90% for identification of behaviors, and Cohen’s kappa ≥ 0.80 for categorical variables). Following this training period, coders continued to double-code 20% of the videos to ensure data quality and prevent coder drift. Reliability videos were selected at random with the constraint that earlier and later sessions were equally represented. Coding discrepancies were discussed and resolved through group consensus. Original codes were used to calculate reliability data, but revised codes were used for data included in final analyses.

Infant Communication.

All deictic gestures produced by infants were coded. We focused only on deictic gestures because they occur frequently and are not produced as part of ritualized routines (such as waving bye-bye; e.g., Bates et al., 1975). Deictic gestures included: a) gives: infant handed an object to a social partner; b) shows: infant held up an object to show a social partner; c) reaches: infant reached with an open hand to request a distal object; d) index finger points: infant pointed to a distal object with an isolated index finger (i.e., the finger does not contact the object); e.) index finger touches: infant touched an object with an isolated index finger. Only gestures that were produced spontaneously were coded (i.e., gestures that were verbally elicited by a caregiver were excluded; Iverson & Goldin-Meadow, 2005). To illustrate, if a caregiver said, “Can you point to teddy?” and then baby pointed, the pointing gesture was not was not coded. Given the relatively low base rates of infant gestures during the 10-minute observations (see Table 1) we collapsed all deictic gestures into a single category in analyses, rather than analyzing the five gesture types separately.

Table 1.

Descriptive statistics (mean, median, range) for variables at all timepoints.

	Walking experience (months)
	−3	−2	−1	0	1	2	3
Infant variables (M, Mdn, range)
Gestures	2.08, 2.00 (0 – 11)	3.44, 2.00 (0 – 16)	4.48, 2.00 (0 – 20)	4.63, 3.00 (0 – 17)	6.40, 5.00 (1 – 17)	7.96, 6.00 (1 – 26)	10.75, 9.00 (0 – 31)
Vocalizations	22.08, 17.50 (2 – 57)	18.48, 16.00 (1 – 46)	28.88, 31.00 (1 – 67)	36.33, 38.50 (10 – 78)	35.48, 35.00 (3 – 77)	42.28, 42.00 (9 – 87)	46.38, 46.50 (17 – 81)
Moving communication	1.33, 0.00 (0 – 11)	2.84, 1.00 (0 – 23)	5.48, 4.00 (0 – 20)	7.88, 4.50 (0 – 41)	10.68, 9.00 (0 – 51)	15.80, 14.00 (0 – 43)	21.96, 22.00 (4 – 51)
Stationary communication	22.83, 18.50 (3 – 59)	19.08, 18.00 (3 – 49)	27.88, 22.00 (2 – 67)	33.08, 34.00 (9 – 67)	31.20, 29 (10 – 69)	34.44, 36.00 (9 – 63)	35.17, 33.00 (8 – 68)
Caregiver variables (M, Mdn, range)
Moving responses	0.33, 0.00 (0 – 3)	1.16, 0.00 (0 – 11)	1.60, 1.00 (0 – 7)	1.88, 1.00 (0 – 18)	5.32, 3.00 (0 – 25)	7.68, 5.00 (0 – 25)	11.88, 13.00 (0 – 39)
Stationary responses	4.04, 2.50 (0 – 20)	6.40, 6.00 (0 – 20)	8.00, 7.00 (0 – 26)	9.46, 8.00 (0 – 22)	12.96, 13.00 (0 – 30)	15.75, 14.00 (3 – 38)	15.21, 14.00 (0 – 34)

Coders also identified infant vocalizations, defined as any pre-speech sound the infant produced. Vegetative sounds (e.g., coughs, sneezes, breathing) and affective sounds (e.g., laughing, fussing, crying) were not coded.

Co-occurrence of Infant Locomotion and Communication.

Coders indicated whether gestures and vocalizations cooccurred with infant locomotion (i.e., were “moving”). Communicative behaviors were coded as moving if they temporally overlapped with a bout of crawling, cruising, or walking, or if the behavior occurred within 2 seconds after the infant stopped moving. The 2-second window allowed us to capture instances when infants locomoted and immediately communicated. All gestures and vocalizations not identified as moving were coded as stationary. Preliminary analyses indicated that gestures and vocalizations were consistent in the extent to which they were moving or stationary. That is, moving gestures and vocalizations showed similar growth patterns; and the same was true for stationary gestures and vocalizations. Thus, in final analyses we collapsed moving behaviors into a single “moving communication” variable, and collapsed stationary behaviors into a “stationary communications” variable.

Caregiver Responses.

Lastly, coders indicated whether each infant gesture and vocalization received a contingent verbal response from a caregiver. A verbal response was identified if the caregiver produced a linguistic vocalization (i.e., containing words) within 2 seconds of the infant gesture or vocalization. These criteria have been previously used to capture patterns of caregivers’ contingent responses (e.g., Gros-Louis, West & Goldstein, 2006; Tamis-LeMonda, Bornstein & Baumwell, 2001).

Inter-rater Reliability.

First, we calculated how closely primary and secondary coders identified infants’ gestures and vocalizations. An agreement was counted if both coders identified the same behavior within one second of each other (i.e., if both coders specified that a “point” had occurred with less than a second separating the behaviors). All other instances were considered disagreements, including when coders categorized the behavior differently (e.g., one coder identified a point and another coder identified a reach), or when only one coder identified a behavior. Mean percent agreement was 86.9% for identification of gestures (gives = 87.3%, shows = 89.1%, reaches = 85.1%, index finger points = 85.8%, index finger touches = 92.0%) and 91% for identification of vocalizations.

In a second pass, after codes had been finalized for identification of gestures and vocalizations, coders watched each behavior and made categorical decisions about whether the behavior was moving or stationary and whether it received a caregiver response. Cohen’s kappas were calculated to compute reliability on categorical coding decisions. Cohen’s kappas were κ = 0.95 for categorizing gestures as moving and κ = 0.98 for categorizing vocalizations as moving. For caregiver response variables, Cohen’s kappas were κ = 0.93 for caregiver verbal responses to gestures, and κ = 0.88 for caregiver verbal responses to vocalizations.

Analytic Approach

The present study measured changes in: 1.) infants’ production of gestures and vocalizations; 2.) coordination of locomotion and communication; and 3.) caregivers’ contingent responses to infant communication across the transition from crawling to walking. Hierarchical linear modeling (HLM) was used to achieve these objectives. HLM partitions the variance of nested data into within-cluster effects and between-cluster effects. This is well-suited for these data, as time-points are nested within individual infants. At the within-infant level (Level 1), we assessed how time-varying factors (e.g., walking experience) accounted for variance in dependent variables. At the between-infant level (Level 2), we assessed how time-invariant factors (age at walk onset) accounted for variance in infants’ intercept and slope terms. An additional advantage of HLM is that it accommodates missing and unequally spaced data (e.g., Huttenlocher, Haight, Bryk, Seltzer & Lyons, 1991; Singer, 1998). For the present study, 168 of 175 (96.0%) observations were completed. Data were analyzed using Version 7 of HLM for Windows (Raudenbush, Bryk, Cheong, Congdon & du Toit, 2011).

Model Selection

Data analysis began by selecting the best-fitting, most parsimonious model for each dependent variable, which was completed through a multistep process. For each dependent variable, four unconditional models—i.e., models with no predictors other than intercept and slope terms—were calculated. The first was an intercept-only model. An intercept-only model may seem counterintuitive for longitudinal data, but it is the most appropriate when the dependent variable remains stable over time (i.e., it depicts a flat line). Second, a linear model with TIME as a predictor was fitted. Next, a quadratic model with TIME and TIME2 predictors was calculated. Finally, a piecewise linear model was fitted, with PIECE 1 TIME and PIECE 2 TIME included as predictors. A piecewise model estimates growth over time as two pieces, rather than a single continuous variable. PIECE 1 estimates baseline growth in the dependent variable across the entire period (i.e., all 7 timepoints in this study). PIECE 2 estimates linear growth following an inflection point (here, the session prior to walk onset; Figure 1 provides a schematic illustration of how time is coded). A significant PIECE 2 slope denotes an incremental change in linear growth following the inflection point.

Figure 1.

Schematic illustration of the Piece 1 and Piece 2 time variables. The top row illustrates how Piece 1 was coded, and represents the baseline linear growth (dotted line). The bottom row illustrates how Piece 2 was coded (solid line). Unlike Piece 1, only months with positive walking experience are included. Piece 2 represents additional incremental linear growth during walking months only.

Next, we compared the fit of each unconditional model by calculating a deviance score for each model, which compares the real observed values against the model-predicted values. Model deviance scores were then compared using chi-square statistics. Higher-order growth models were selected over simpler models only if they significantly reduced deviance, and if the additional growth term was significantly greater than zero. In some instances, piecewise and quadratic models did not significantly differ in their model fit. When this occurred, the model with the lower deviance score was selected. This process led us to select linear and piecewise models to describe growth trajectories.

Final Conditional Models

Linear Models.

A linear model was selected as the best fit to model infants’ vocalizations, stationary communications, and caregiver responses to infant communications. For these linear models, Level 1 estimated individual linear growth across the period as a function of TIME. We centered the data at the mid-point: the visit prior to walk onset. This point was selected because it marks the very start of the transition to walking. The equation for Level 1 is:

$${\text{Y}}_{\text{ti}}={\text{Π}}_{0\text{i}}+{\text{Π}}_{1\text{i}}\ast \left({\text{TIME}}_{\text{ti}}\right)+e_{\text{ti}}$$ (1)

Here the intercept (π_0i) represents the level of the dependent variable of infant i at the midpoint (e.g., the estimated number of vocalizations an infant produced at the visit prior to walk onset). The term π1i represents the linear slope—the rate and direction of change across the period—for infant i.

At Level 2, infants’ age at walk onset was included as a predictor on intercept and slope terms. This allowed us to assess whether infants’ chronological age influenced their growth trajectory over this transition. The final Level 2 equations for the final linear model were as follows:

$${\pi }_{0\text{i}}={\mathrm{\beta }}_{00}+{\mathrm{\beta }}_{01}^{\star }\left({\text{ Age at Walk Onset }}_{\text{i}}\right)+{\text{r}}_{0\text{i}}$$ (2)

$${\pi }_{1i}={\mathrm{\beta }}_{10}+{\mathrm{\beta }}_{11}\ast \left({\text{ Age at Walk Onset }}_{\text{i}}\right)+{\text{r}}_{1\text{i}}$$ (3)

Here, coefficients β₀₁ and β₁₁ represent the effect that a one-month increase in the age at walk onset has on infants’ intercept and slope terms.

Piecewise Models.

A piecewise model was the best fit for modeling gestures, moving communication, caregiver responses to moving communications. In these piecewise models, growth was estimated as a function of two slope terms: PIECE 1 slope (all time points; baseline growth) and PIECE 2 slope (incremental growth after walk onset). Again, we centered the data at the midpoint, the visit prior to walk onset. The equation for Level 1 is as follows:

$${\text{Y}}_{\text{ti}}={\text{Π}}_{0\text{i}}+{\text{Π}}_{1\text{i}}\ast \left(\text{PIECE}{1}_{\text{t}}\right)+{\text{Π}}_{2\text{i}}\ast \left(\text{PIECE}{2}_{\text{ti}}\right)+e_{\text{t}}$$ (1)

Again, the intercept (π0i) represented infant i’s score at the visit prior to walk onset (e.g., how many gestures infants produced at the visit before walk onset). The Piece 1 slope represents the estimated baseline linear growth rate for infant i (e.g., the growth in gestures over the entire period), and the Piece 2 slope represents the estimated additional incremental growth from the visit prior to walk onset forward for infant i (e.g., the growth in gestures after the final crawling session). See Figure 1 for a depiction of the coding of piecewise time.

Level 2 predictors were consistent with the linear models, and the equations are as follows:

$${\pi }_{0i}={\mathrm{\beta }}_{00}+{\mathrm{\beta }}_{01}\ast \left({\text{ Age at Walk Onset }}_{\text{i}}\right)+{\text{r}}_{0i}$$ (2)

$${\pi }_{1\text{i}}={\mathrm{\beta }}_{10}+{\mathrm{\beta }}_{11}\ast \left({\text{ Age at Walk Onset }}_{\text{i}}\right)+{\text{r}}_{1\text{i}}$$ (3)

$${\pi }_{1\text{i}}={\mathrm{\beta }}_{20}+{\mathrm{\beta }}_{21}\ast \left({\text{ Age at Walk Onset }}_{\text{i}}\right)+{\text{r}}_{2\text{i}}$$ (4)

As in previous models, the β terms allow us to examine the effect of age at walk onset on the model terms (the intercept, Piece 1 slope, and Piece 2 slope).

Results

Our objective was to document changes in infant communication and caregiver responses across the transition from crawling to walking. Specifically, we investigated whether the onset of walking was an inflection point—a point in the developmental trajectory when growth changed substantially—in communicative development. We report data on how frequently infants gestured, vocalized, paired communicative behaviors with locomotion, and on caregivers’ contingent verbal responses to infant communications.

Descriptive statistics for all variables are shown in Table 1. As can be seen in the table, infant communications increased over time. Gestures increased modestly during the crawling months: from M = 2.08 (Mdn = 2.00, range = 0–11) at the first crawling session, to M = 4.63 (Mdn = 3.00, range = 0–17) at the final crawling session. However, gesture growth increased substantially during the walking months—increasing to M = 10.75 (Mdn = 9.00, range = 0–31) gestures by the final walking session. Vocalizations increased steadily across sessions, from M = 22.08 (Mdn = 17.50, range = 2–57) initially to M = 46.38 (Mdn = 46.50, range = 17–81) vocalizations at the final session.

Both moving and stationary communications increased over time, but they showed different growth patterns. Stationary communications increased steadily across the entire period, from M = 22.83 (Mdn = 18.50, range = 3–59) initially to M = 35.17 (Mdn = 33.00, range = 8–68) at the final session. Conversely, infants’ moving communications increased modestly during the crawling sessions—from M = 1.33 (Mdn = 0.00, range = 0–11) to M = 7.88 (Mdn = 4.50, range = 0–41)—but experienced dramatic growth during walking months, increasing to M = 21.96 (Mdn = 22.00, range = 4–51) at the final session.

Caregiver responses to moving and stationary communications followed much the same pattern. Responses to stationary communication increased at a steady pace from the first (M = 4.04, Mdn = 2.50, range = 0–20) to last session (M = 15.21, Mdn = 14.00, range = 0–34). Responses to moving communication remained at low-levels during crawling months—from M = 0.33 (Mdn = 0.00, range = 0–3) to M = 1.88 (Mdn = 1.00, range = 0–18) at first and final crawling months—and dramatically increased when infants began walking, increasing to M = 11.88 (Mdn = 13.00, range = 0–39) at the last session. Final model coefficients are presented in Table 2 for infant variables, and Table 3 for caregiver variables.

Table 2.

Final models predicting growth trajectories in variables with Walk Onset Age and Sex as predictors of the slope and intercept terms;

	Gestures		Vocalizations		Moving Communication		Stationary Communication
	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Intercept, β00	4.77***	0.63	32.82***	1.96	7.23***	1.54	29.08***	1.46
Walk onset age, β01	1.38**	0.44	1.55	1.01	0.30	0.71	1.74*	3.09
Sex, β02	0.47	1.23	−1.86	4.05	−0.94	2.73	−1.89	3.09
Linear slope, β10	0.81***	0.20	4.55***	0.59	2.05**	0.62	2.55***	0.48
Walk onset age, β11	0.16	0.16	1.05*	0.45	−0.19	0.35	0.88**	0.30
Sex, β12	−0.03	0.39	3.41**	1.17	0.52	1.11	2.03	1.02
Piecewise slope, β20	1.04*	0.49	--	--	2.49*	1.08	--	--
Walk onset age, β21	0.11	0.30			1.11*	0.53
Sex, β22	0.70	0.98			2.48	1.94

^*p < .05,

^**p < .01,

^***p < .001

Table 3.

Final models predicting growth trajectories in caregiver variables with infants’ Walk Onset Age and Sex as predictors of the slope and intercept terms;

	Responses to moving communication		Responses to stationary communication
	Coeff.	S.E.	Coeff.	S.E.
Intercept, β00	1.99**	0.66	10.29***	0.60
Walk onset age, β01	0.17	0.27	2.13***	0.44
Sex, β02	0.35	1.17	−0.61	1.27
Linear slope, β10	0.52	0.26	2.07***	0.23
Walk onset age, β11	−0.06	0.14	0.51***	0.11
Sex, β12	0.48	0.48	0.80	0.50
Piecewise slope, β20	2.60***	0.67	--	--
Walk onset age, β21	0.75**	0.26
Sex, β22	1.35	1.27

^*p < .05,

^**p < .01,

^***p < .001

Infant Communication

Gestures.

A piecewise model best fit infants’ gesture growth. Model estimates are graphed in Figure 2A (left), showing that infants’ gesture production increased linearly over time, with additional growth after walk onset. During crawling months, gesture production increased at an estimated rate of 0.81 gestures per month (β₁₀ = 0.81, SE = 0.20, p < 0.001). During walking months, production increased at a rate of 1.85 gestures per month, more than doubling the pre-walk rate (0.81 gestures per month baseline growth + 1.04 gestures per month additional incremental growth; β₂₀ = 1.04, SE = 0.49, p = 0.047). On the right panel of Figure 2A, we plotted how each infants’ slope changed at walk onset (i.e., positive values indicate that growth increased; negative values indicate that growth decreased). The majority of infants (20 of 25) displayed a pattern of additional growth at walk onset, with many showing a dramatic increase.

Figure 2.

Growth trajectories of infant gestures and vocalizations. (A) Infant gestures. (B) Infant vocalizations. The left panels depict the modeled growth in gestures (top) and vocalizations (bottom), with the 95% confidence interval shaded in grey. The vertical dotted line marks the final crawling-only visit. On the right panels, each infants’ change in slope at walk onset is plotted, where positive values reflect an increase in slope at walk onset, and negative values reflect a decrease. The color of each infants’ dot denotes their age at walk onset.

Older infants gestured more than younger infants did at the intercept (i.e., the final crawling session; β₀₁ = 1.38, SE = 0.45, p = 0.005). However, age at Walk Onset had no effect on the slope terms, indicating that infants experienced additional growth in gestures following walk onset regardless of when they began walking.

Vocalizations.

A linear model best fit infants’ vocalization growth. Model estimates are graphed in Figure 2B (left), showing that infants’ vocalizations grew linearly over time, increasing by 4.55 vocalizations/month (β₀₁ = 4.55, SE = 0.60, p < 0.001). Thus, infants nearly doubled their volubility from the first to final session, increasing from an estimated 19.16 to 46.47 vocalizations. The fact that a linear model best fit the data indicates that growth was stable across the entire observation period and did not change at walk onset. This is further supported by inspection of individual data. We plotted how individual infants’ slopes changed at walk onset in Figure 2B. Only a minority of infants’ slopes increased at walk onset. In fact, 16 infants either showed no change or a decline in their growth rate at walk onset.

The model also revealed that older infants tended to show steeper growth in vocalizations over time (β₁₁ = 1.05, SE = 0.45, p = 0.03). A one month increase in Walk Onset Age was associated with an additional 1.05 vocalizations added per month.

Coordination of Infant Locomotion and Communication

Moving communication.

A piecewise model best fit trajectories of infants’ moving communications (i.e., gestures and vocalizations that were coordinated with locomotion). Model estimates are graphed in Figure 3A, showing that infants’ moving communications increased linearly over time, with additional growth after walk onset. During crawling months, infants increased by an estimated 2.05 moving communications each month (β₁₀ = 2.05, SE = 0.62, p = 0.003). During walking months, infants increased by 4.54 moving communications each month (2.05 baseline growth plus an additional 2.49 behaviors per month; β₂₀ = 2.49, SE = 1.08, p = 0.03). The right panel of Figure 3A illustrates each infants’ change in slope at walk onset, showing that 21 of 25 infants increased growth in moving communication following walk onset.

Figure 3.

Growth trajectories of infant’ moving and stationary communications. (A) Moving communications. (B) Stationary communications. The left panels depict the modeled growth in moving (top) and stationary (bottom) communication, with the 95% confidence interval shaded in grey. The vertical dotted line marks the final crawling-only visit. On the right panels, each infants’ change in slope at walk onset is plotted, where positive values reflect an increase in slope at walk onset, and negative values reflect a decrease. The color of each infants’ dot denotes their age at walk onset.

Walk Onset Age had no effect on infants’ intercepts or baseline slopes. However, it did account for variance in piece 2 slopes (i.e., the additional growth after walk onset; β₂₁ = 1.11, SE = 0.53, p = 0.049). A one month increase in Walk Onset Age was associated with an additional 1.11 moving communications per month. That is, infants who began to walk later in development showed greater gains in moving communication growth than did younger infants.

Stationary communication.

A linear model best fit trajectories of infants’ stationary communication (i.e., gestures and vocalizations that were not coordinated with locomotion). Model estimates are graphed in Figure 3B, showing that infants’ stationary communications grew linearly over time, increasing by 2.55 behaviors per month (β₁₀ = 2.55, SE = 0.48, p < 0.001).

Additionally, older infants produced more stationary communications at the intercept (β₀₁ = 1.74, SE = 0.70, p = 0.021) and showed steeper growth over time compared to younger infants (β₁₁ = 0.88, SE = 0.30, p = 0.008). A one-month increase in age at walk onset was associated with 1.74 more stationary communications at the intercept, and an additional 0.88 stationary communications per month.

Caregiver Responses to Communication

Responses to moving communication.

A piecewise model best fit trajectories of caregiver responses to infants’ moving communications. Model estimates are graphed in Figure 4A, showing that caregiver responses remained flat during crawling months, but there was linear growth after infants started walking. During crawling months, caregiver responses to moving communications remained consistently low, with no significant growth over time (β₁₀ = 0.52, SE = 0.27, p = 0.065). However, during walking months, caregiver responses showed positive linear growth, increasing by 3.12 responses each month (β₂₀ = 2.60, SE = 0.67, p < 0.001). The right panel of Figure 4A plots each caregivers’ change in slope when infants began walking, revealing consistency among caregivers: 22 of 25 caregivers displayed increased growth in responses to moving communications when infants started walking.

Figure 4.

Growth trajectories of caregiver responses to infants’ moving and stationary communication. (A) Caregiver responses to moving communication. (B) Caregiver responses to stationary communication. The left panels depict the modeled growth in responses to moving (top) and stationary communications (bottom), with the 95% confidence interval shaded in grey. The vertical dotted line marks the final crawling-only visit. On the right panels, each caregivers’ change in slope at walk onset is plotted, where positive values reflect an increase in slope at walk onset, and negative values reflect a decrease. The color of each dot denotes the infants’ age at walk onset.

Notably, caregivers of older infants showed steeper growth in their responses to moving communication after infants began to walk, compared to caregivers of younger infants (β₂₁ = 0.75 SE = 0.26, p = 0.010). A one month increase in infants’ age at walk onset was associated with an additional 0.75 responses to moving communications per month.

Responses to stationary communication.

A linear model best fit trajectories of caregiver responses to stationary communication. Model estimates are graphed in Figure 4B, showing that caregiver responses to stationary communication grew linearly over time, increasing by 2.07 responses per month (β₁₀ = 2.07, SE = 0.23, p < 0.001).

Compared to caregivers of younger infants, caregivers of older infants responded to more stationary communications at the intercept (β₀₁ = 2.13, SE = 0.44, p < 0.001) and showed steeper growth over time (β₁₁ = 0.51, SE = 0.11, p < 0.001). A one-month increase in infants’ age at walk onset was associated with 2.13 more responses to stationary communications at the intercept, and an additional 0.51 responses per month.

Discussion

The notion that motor development influences other psychological domains has deep roots in infant research (e.g., Campos et al., 2000; Gibson, 1988; Piaget, 1954). Esther Thelen noted in 2000 that, “movement helps children sample the world more completely” (p. 394)—this surely applies to walking. Walking enables infants to move faster and farther than crawling, and consequently expands their’ “sampling” of the visual, spatial, and social information available in the environment (see for a discussion: Adolph & Tamis-LeMonda, 2014). Does this developmental cascade extend to infants’ communication? We found that after infants began walking, the pace of gesture growth increased substantially, and infants increasingly coordinated gestures and vocalizations with locomotion (e.g., by walking to a caregiver and showing off a toy bear). Consequently, caregivers had more opportunities to respond to infants during walking months compared to crawling months. Additionally, changes in communication were amplified for infants who walked later in development, compared with younger walkers.

Infants coordinate communication behaviors with walking

Infants gestured more frequently during walking than crawling sessions, replicating prior findings (Clearfield, 2008; Walle, 2016). Importantly, our data also illuminated an inflection point in gesture development: the pace of gesture growth more than doubled after walk onset. Relative to when they were only crawling, infants pointed, gave and showed toys to caregivers, and requested objects more frequently after they began walking. Although we cannot draw a causal link from observational data, the uptick in gesture growth is consistent with a developmental cascades perspective, namely that walking provides increased opportunities for infants to gesture. Prior work by Karasik and colleagues suggests a potential mechanism for this increase. They found that walkers carried objects while locomoting more often than same-aged crawlers. Moreover, crawlers’ hands were typically multi-tasking while carrying objects: that is, their hands held the object while simultaneously supporting their posture. Increased object carrying may enable walkers to produce more frequent gestures—namely, showing and giving objects to caregivers—compared to crawlers (Karasik, Adolph, Tamis-LeMonda & Zuckerman, 2012).

Infant vocalizations increased steadily over time, but there was no sharp change at walk onset. Why might gestures, but not vocalizations, increase at walk onset? Unlike vocalizations, gestures are visual cues, and caregivers need to see infants to notice their points, requests, and shows. Walking allows infants to easily approach caregivers—entering caregivers’ visual field—and presumably, this makes infant gestures more visible to caregivers than if the infant had gestured from a distant location. Conversely, vocalizations are an audible cue. Caregivers can notice vocalizations (and therefore respond) even when not looking directly at the infant. Indeed, infants tend to vocalize just as often when caregivers are located close or far away (Anderson, Vietze & Dokecki, 1978).

After walk onset, infants increasingly produced moving communications (e.g., by approaching a caregiver and gesturing). It is possible that locomotion and communication undergo simultaneous, but unrelated, gains during development (i.e., both skills are driven by general maturational change). However, the fact that we observed a dramatic increase in moving communication, but not stationary communication, suggests otherwise. If maturational change alone accounted for communication growth, we would expect communication overall—not just moving communication—to undergo significant growth at this time. The finding that moving communication alone changed at walk onset suggests that the ability to walk feeds into infants’ moment-to-moment communicative behaviors.

Of course, locomotion and communication are likely related in bidirectional, complex ways throughout infancy. For instance, the process of learning to walk is presumably influenced by infants’ social environments. Caregivers can encourage locomotion by calling infants over or directing them to retrieve far-off objects. Caregivers can also constrain locomotion by placing infants in high-chairs, play pens, and car seats that prevent movement. There are likely many rich and elaborate interconnections among infants’ walking, exploration, and social-communication environment beyond those measured here.

Walking has a cascading effect on caregiver responses

As infant communication changed over time, we observed corresponding changes in caregivers’ contingent responses. Consistent with prior work, caregivers responded to a greater proportion of infants’ moving communications compared to stationary communications (Karasik, Tamis-LeMonda & Adolph, 2014; Toyama, 2018). And, following walk onset, caregivers had substantially more opportunities to respond to moving bids. Why do moving communications receive more responses than stationary communications? Infants’ locomotion to caregivers puts the dyad in close proximity, making infant behaviors more visible. In addition, the multimodal coordination of locomotion and communication is likely salient to caregivers, perhaps denoting a high degree of persistence and directedness from infants.

Prior work also indicates that the content of caregiver responses differs for moving versus stationary communication. A study by Karasik et al (2014) found that caregivers frequently provided actions directives (e.g., “stack it”, “go give it to Daddy”) to infants’ moving bids, but used more referential language (e.g., “that’s a spoon”) and affirmations (e.g., “thank you”) when responding to stationary bids. Different responses could, in turn, reciprocally affect infants’ locomotion (e.g., caregivers’ action directives may prompt new infant actions).

Collectively, our findings and prior work highlight qualitative changes in caregivers’ language input when infants begin to walk. Such changes may help to explain prior findings by Walle and Campos (2014), who reported that caregivers directed similar amounts of language input to crawling and walking infants—but importantly, caregiver input strongly predicted vocabulary size for walking infants, but not for crawling infants. The onset of walking may give rise to moments where language input is optimal for word learning; for instance, bouts of contingent responses to infant bids (e.g., “did you find teddy?” as infant approaches with teddy bear) and action directives that align with the infants’ ongoing action (e.g., “bring it to me!” as an infant carries a toy). Indeed, prior work suggests that both receptive and expressive vocabulary show increased growth when infants begin walking (He, Walle & Campos, 2016; Walle & Campos, 2014; West, Leezenbaum, Northrup & Iverson, 2018; however, Moore et al., 2019, find no such growth). Increased growth in infants’ expressive and receptive vocabulary following walk onset may feed back into the quality of caregivers’ responses.

Changes in caregiver responses at walk onset may spur changes in other domains of infant development. The back-and-forth interactions in infant-caregiver dyads provide rich material for infant learning, including word learning, advanced play skills, and prosocial behavior (e.g., Bornstein & Tamis–LeMonda, 1989; Newton et al., 2014; Nicely, Tamis-LeMonda & Bornstein, 1999). Indeed, the increase in caregiver responsivity could, in part, account for findings that word learning accelerates when infants begin to walk (e.g., He, Walle & Campos, 2016; Walle & Campos, 2014; West et al., 2019).

Timing matters: Later walkers exhibit greater changes

Infants began walking at different ages, ranging from 9- to 15-months-old. Regardless of age, walk onset consistently co-occurred with increased gestures, moving communications, and caregiver responses to moving communication. However, the magnitude of the increase was greater for older than for younger walkers. The finding may be explained by infants’ prior communicative development. Infants bring their prior experiences and past accomplishments to the table when they begin to walk. Likely, walking provides opportunities for infants to showcase communicative behaviors already in their repertoires, and older infants may have more sophisticated communicative repertoires than younger infants. Indeed, the period from 9- to 15-months-old is marked by substantial change in infants’ ability to gesture, vocalize, and eventually even produce words (e.g., Bates, 1976; Bates, Benigni, Bretherton,Camaioni, & Volterra, 1979; Iverson et al., 2018).

Notably, Biringen and colleagues (1995) reported that younger infants showed a greater increase in their emotional expressions (both positive and negative) at walk onset compared to older infants. Taken together with our findings, it seems possible that infants’ predominant communicative behaviors increase with walking—for younger infants, the predominant behaviors may be emotional affectations (e.g., fussing to request an out-of-reach bottle) and for older infants, predominant behaviors are gestures (e.g., pointing to request an out-of-reach bottle).

It is also worth noting that our sample began walking within a typical age range, defined as 8.2 to 17.6 months by the World Health Organization (WHO Multicentre Growth Reference Study Group & de Onis, 2006). Future work is needed to determine if infants who walk outside of this range experience a different pattern of communicative growth. In particular, studies of infants with Type 2 spinal muscular atrophy may be especially informative, as these infants experience extreme deprivation in locomotor experiences but demonstrate typical communicative development as measured by the MacArthur-Bates Communicative Development Inventory (e.g., Sieratzki & Woll, 2002; Fenson et al., 1994). Although we hypothesize that walking is typically participatory in the development of communication, this does not preclude other, alternative developmental pathways (i.e., walking is clearly not necessary or sufficient for advances in communication).

Limitations

This study has notable strengths, including longitudinal data acquired over seven months that allowed us to establish baseline growth in each variable and test for an inflection point at walk onset. However, there are also some important limitations. First, the study design was observational and therefore we cannot establish a causal link between walking and communication. And indeed, the links between walking and communication are likely bidirectional. Moreover, other unmeasured variables likely influence both walking and communication. For example, features of infants’ temperament are related to language development (e.g., Slomkowski, Nelson, Dunn & Plomin, 1992), and the timing of motor skill attainment (Biringen et al., 2008). Future studies with experimental designs—in which researchers manipulate infant locomotion and observe the effect on communication—could help to disentangle the effect of infants’ real-time locomotion on their communication (e.g., Clearfield, 2011).

Additionally, we sampled 10 minutes of behavior at each session, which may have limited our ability to capture behaviors with very low base rates. In particular, we were unable to measure infants’ language production because infants in our sample rarely uttered words. At this point in development—around the end of the first year—infants are just beginning to produce words (e.g., Fenson et al., 1994). Longer observations may have yielded more utterances containing words, and provided an opportunity to measure infants’ spontaneous language production.

Third, our observations focused specifically on play interactions (caregivers were instructed to “play as you typically would”). However, caregivers may structure play activities differently for crawling and walking infants during daily routines. Likewise, crawling and walking infants may differ in how frequently or effectively they instigate play with caregivers. Future research should replicate findings with observations of infants’ daily routines, which likely include activities like meal-times and chores in addition to play.

Finally, our sample included predominantly white, highly educated families. Additional work is necessary to determine whether these findings replicate among families from different racial, ethnic, and socio-economic backgrounds.

Conclusions

Our findings provide compelling support for a developmental cascades hypothesis—that learning to walk enriches infants’ communicative environment. When infants begin to walk, they have greater autonomy to travel to distant places, objects, and people in their environment than they did as crawlers. Walking infants leverage this autonomy by approaching caregivers to gesture or bid for attention. Thus, infant communication becomes more frequent, targeted, and salient to caregivers, who in turn provide more frequent responses to their infants. Changes in the back-and-forth everyday interactions between infants and caregivers likely further shape opportunities for infant learning in other higher-level domains. Finally, this work underscores the value of a multi-disciplinary approach to understand infants’ social-communicative development within the context of their dramatically-changing repertoires for motor action.

Research Highlights.

• After infants began to walk, the pace of their gesture growth (but not vocalization growth) increased substantially.

• Infants increasingly paired their communication behaviors with locomotion after learning to walk.

• Changes in infant communication were accompanied by changes in caregivers’ verbal responses to infants.

• Changes in communication were amplified for infants who began to walk at older ages compared to infants who walked earlier.