|Home | About | Journals | Submit | Contact Us | Français|
In uncertain situations such as descending challenging slopes, social signals from caregivers can provide infants with important information for guiding action. Previous work showed that 18- month-old walking infants use social information selectively, only when risk of falling is uncertain. Experiment 1 was designed to alter infants’ region of uncertainty for walking down slopes. Slippery Teflon-soled shoes drastically impaired 18-month-olds’ ability to walk down slopes compared with walking barefoot or in standard crepe-soled shoes, shifting the region of uncertainty to shallower range of slopes. In Experiment 2, infants wore Teflon-soled shoes while walking down slopes as their mothers encouraged and discouraged them from walking. Infants relied on social information on shallow slopes, even at 0°, where the probability of walking successfully was uncertain in the Teflon-soled shoes. Findings indicate that infants’ use of social information is dynamically attuned to situational factors and the state of their current abilities.
In situations of uncertainty, infants often look to their caregivers to decide how to respond. Is the stranger friend or foe? Is the new toy funny or scary? Is the unfamiliar room inviting or threatening? Parents’ social signals can tell infants whether to approach or avoid a new person, novel object, or questionable situation.
In the most commonly used paradigms in the social referencing literature, researchers observe infants’ responses to an “ambiguous” toy (e.g., a motorized robot) or a strange person (Baldwin & Moses, 1996; Feinman, Roberts, Hsieh, Sawyer, & Swanson, 1992; Stenberg, 2009; Stenberg & Hagekull, 2007). Infants solicit information from their mothers by looking toward them or engaging them in social interaction. Mothers, in turn, provide infants with social information about how to interpret the stimulus by varying their facial expressions and vocalizations. By 12 months of age, infants change their affective displays and proximity to mother, toy, or person in accordance with the valence of mothers’ messages (Hirschberg & Svejda, 1990; Hornik, Risenhoover, & Gunnar, 1987; Mumme, Fernald, & Herrera, 1996; Stenberg, 2009; Stenberg & Hagekull, 1997).
Although infant locomotion may seem far afield from social cognition, locomotor tasks are especially useful for studying infants’ sensitivity to social information. In locomotor tasks, researchers can parametrically vary the level of risk and ambiguity, whereas in other paradigms, researchers must rely on their own intuitions about ambiguity to select toys and strange people. Moreover, obstacles to locomotion elicit particularly clear approach/avoidance responses accompanied by a variety of social expressions and exploratory behaviors (Tamis-LeMonda & Adolph, 2005). Locomotion itself is a compelling attractor and infants crawl or walk for the sheer joy of moving (Harlow & Mears, 1979). In the laboratory, they happily clamber up and down obstacles during dozens of trials for upwards of 60 minutes. At the same time, risk of falling is a powerful deterrent so infants are highly motivated to guide locomotion adaptively. They consistently avoid obstacles that pose a clear threat of falling (Adolph, 1997).
Indeed, the most famous laboratory demonstration of the power of social information for guiding action is Sorce et al.’s (1985) study of mothers’ emotional signaling to their 12-month-olds on a modified “visual cliff.” In the standard paradigm, infants face their mothers from one side of a glass-covered table (Walk & Gibson, 1961). On the “shallow” side of the apparatus, a patterned surface is attached to the underside of the safety glass so that it appears solid. On the “deep” side, the patterned surface is 102 cm below the glass, presenting the illusion of a large drop-off. In Sorce et al.’s study, the apparent drop-off was set to 30 cm, a height presumed to be ambiguous. Mothers stood at the far side of the obstacle and displayed positive and negative facial expressions. Infants who looked toward their mothers’ faces were swayed by the emotional signals: They crossed when mothers displayed positive expressions and avoided when mothers displayed negative expressions. In a subsequent study, 12-month-olds were faster to cross the visual cliff in response to mothers’ vocal encouragement than to her positive facial expressions (in the voice-only condition, mothers stood with their backs to the apparatus) and fastest when they could both hear and see mother, suggesting that the voice is an especially important channel for social signals (Vaish & Striano, 2004).
Of course, infants don’t always seek parents’ counsel or take their unsolicited advice. When crawling infants are tested on the shallow side of the visual cliff, mothers’ negative facial expressions are superfluous. Most 12-month-olds crossed without ever glancing at their mothers’ faces, and those who view mothers’ fearful faces were not deterred (Sorce, et al., 1985). Likewise, encouragement is equally ineffective at swaying infants’ responses on the deep side of the cliff; in the standard set-up with a 90-cm drop-off, 12-month-old crawlers avoided crossing the deep side despite mothers’ smiles and verbal coaxing (Richards & Rader, 1983; Walk, 1966; Witherington, Campos, Anderson, Lejeune, & Seah, 2005).
More subtle variation of risk level reveals exquisitely selective use of social information for guiding locomotion down slopes (Tamis-LeMonda, Adolph, Lobo, Karasik, & Dimitropoulou, 2008). By 18 months of age, most infants are experienced walkers (infants begin walking around 12 months), but their skill varies widely. Some infants can walk down steep slopes (30°–40°), but others can manage only relatively shallow ones (10°–20°). Thus, a 25° slope can be safe for a more proficient walker (probability of success is 1.0) and impossible for a less skilled walker (probability of success is 0); but for an infant whose walking threshold is around 25°, the risk is uncertain (probability of success is approximately 0.50). By determining each infant’s walking threshold—that is, their level of walking skill—risk can be normalized to skill level. Despite mothers’ encouraging or discouraging facial expressions, gestures, words, and vocal intonations, 18-month-olds ignored mothers’ social signals when slopes were safe or impossible: They walked down safe slopes within their abilities but not down impossible slopes beyond their abilities (Figure 1A). In contrast, on uncertain slopes around their individualized thresholds, infants deferred to mothers’ unsolicited social messages: They walked on 74% of trials when she said go but on only 27% of trials when she said no. The findings indicate that infants weigh social signals provided by their caregivers against perceptual information generated by their own exploratory activity, and social information outweighs perceptual information only when infants believe the latter is ambiguous. The fact that infants responded to mothers’ social signals only at the walking threshold indicates that they correctly recognized the uncertain risk at those increments.
In the literature on locomotor development, Sorce et al.’s (Sorce, et al., 1985) visual cliff study and Tamis-LeMonda et al.’s (2008) slope study are exceptions: The tradition in research on infant locomotion is to ignore effects of social information. Typically, the goal is to assess infants’ movements, rather than their responses to social information. Thus, caregivers merely function as a lure positioned at the far side of a walkway or obstacle, encouraging their infants to crawl or walk. The social context supports the task—infants are more likely to participate with caregivers in view—but since social information is held constant, researchers do not explicitly consider the role of caregivers’ social signals in guiding motor actions. The consequence to the field is a conception of motor development as a solitary enterprise rather than a socially embedded process. More practically, researchers have failed to exploit caregivers’ social signals as a potentially useful and readily available tool for investigating the development of infant locomotion. Differential responding to social information can give researchers purchase on infants’ perception of affordances and their appreciation of motor risk.
An illustration of how manipulation of caregivers’ social signals can inform research on infant locomotion concerns the specificity of motor experience—whether learning transfers from earlier to later developing postures. Several lines of evidence suggest that learning is specific to each posture in development. Novice infants faced with slopes, gaps, and real cliffs in unfamiliar sitting, crawling, cruising, and walking postures attempted impossible increments on repeated trials, requiring rescue by an experimenter to prevent injury from falling. After several weeks of experience with a new posture, infants distinguished safe from impossible increments and scaled their attempts to the conditional probability of success (Adolph, 1997, 2000; Adolph, Tamis-LeMonda, Ishak, Karasik, & Lobo, 2008; Kretch, Karasik, & Adolph, 2009). However, in each of these studies, caregivers provided only encouraging information, resulting in high attempt rates regardless of risk level in the novice infants. In other words, mothers’ encouragement to descend may have led novices to discount the perceptual information specifying risk. But if we tested novice infants with both encouraging and discouraging social signals, we might be better able to detect their sensitivity to risk.
To test this possibility, 12-month-old experienced crawlers and novice walkers were observed on a range of safe, uncertain, and impossible slopes (Adolph, et al., 2008). On half of the trials, mothers encouraged infants to descend and on the other half, mothers discouraged them (as in Tamis-LeMonda, et al., 2008). On impossible slopes, experienced crawlers ignored mothers’ advice; attempt rates were close to 0 in both social conditions. However, social signals affected how novice walkers responded to impossible slopes such that, at the steepest increments, attempt rates decreased but were still perilously high (Figure 1B). Despite mothers’ discouraging frowns and knit eyebrows, prohibitive hand gestures and head shakes, and shouted “No’s” and stern warnings to stop, infants attempted to walk on half of the trials. Thus, it is fair to conclude that under encouraging conditions, novice walkers were not merely walking because of their mothers’ positive social message. Manipulating social signals tells us that novice walkers do have an inkling that extreme increments pose a potential threat of falling, otherwise mothers’ discouragement would have no effect on decisions at all, but they cannot yet scale their perceptual judgments to the limits of their own abilities.
The current experiments took a two-pronged approach to the relations between social cognition and infant locomotion. We designed a new locomotor task to address the selectivity of infants’ social cognition and we manipulated caregivers’ social signals to examine flexibility in infant locomotion. Both approaches were integral to Experiment 2. Experiment 1 merely set the stage by establishing infants’ walking skill under various friction conditions. The idea was to observe experienced 18-month-old walking infants in a slope task similar to the one used by Tamis-LeMonda et al (2008) and Adolph et al (2008), but to experimentally alter infants’ ability to walk so that formerly uncertain slopes (between 20° and 30°) would become impossible and formerly safe slopes (less than 15°) would become uncertain. Experiment 1 established the utility of Teflon-soled shoes to pull off this feat, by pulling infants off their feet, so to speak, with the slippery shoes on carpeted slopes. In Experiment 2, 18-month-olds wore the Teflon-soled shoes and their mothers encouraged and discouraged them from walking down slopes spanning the range from shallow to steep (0° to 50°).
The Teflon-soled shoes allowed us to address the question of whether 18-month-olds can update their assessment of uncertainty and adapt their use of social information accordingly. This question is important because infants’ abilities and the properties of the environment are continually changing. An uncertain slope one week can be safe the next when walking skill improves. A safe slope under one set of conditions (e.g., walking barefoot) can be uncertain or impossible under a different set of conditions (walking in slippery shoes). Keeping pace with such changes in affordances requires real-time updating of risk and dynamically attuned use of social information.
Simultaneously, we manipulated mothers’ social signals—mothers both encouraged and discouraged infants to walk down slopes—to examine two outstanding issues in the development of infant locomotion. One issue concerns the effects of locomotor experience in learning to perceive affordances. Normally, walking experience and walking skill are highly correlated such that more experienced infants display higher walking thresholds than less experienced infants (Adolph, 1997; Tamis-LeMonda, et al., 2008). Thus, an alternative explanation for high error rates in novice walkers on impossible increments is that poor skill overtaxed or interfered with their ability to detect affordances; while trying to keep balance in their new, upright posture, they were unable to focus attention on perceptual information for gauging risk. If we could transform experienced, proficient 18-month-old walkers into poorly skilled walkers while keeping the upright posture constant, then we could disentangle the effect of walking experience from that of walking skill.
A second issue concerns infants’ sensitivity to friction underfoot. Several previous experiments indicated that experienced 14- and 15-month-old walking infants respond only to slant, not to friction as they approach impossible slopes, whereas adults use information for slip underfoot in guiding locomotion down slopes (Adolph, Joh, & Eppler, in press; Joh, Adolph, Narayanan, & Dietz, 2007). But, in previous work, the infants were tested only under encouraging conditions. If we tested infants with both encouraging and discouraging social signals, we might better detect their sensitivity to friction as an important component in affordances for walking down slopes.
Assuming a decrement in infants’ walking skill due to the slippery shoes, several results were possible. Despite several months of walking experience, 18-month-old experienced walkers might behave more like 12-month-old novice walkers and use social signals selectively only at the steepest increments (see Figure 1). Such a finding would suggest that the reason infants display high error rates on impossible increments in novice postures is due to low levels of motor skill, interfering with their ability to focus attention on perceptual information. In this case, errors would not be explained by insufficient walking experience, and prior conclusions about novice infants’ inability to detect affordances would need to be revisited. A second possibility is that the 18-month-olds would respond only to the degree of slant, using social signals selectively only at the mid-range of slopes, where their abilities would be uncertain while walking barefoot or in normal shoes. This result would suggest that infants do not dynamically update their assessment of uncertainty and it would corroborate previous findings that infants are insensitive to variations in friction beneath their feet. A third possibility is that the infants would use mothers’ social signals selectively only at the shallowest range of slopes. This finding would be most exciting because it would suggest that 18-month-olds dynamically update their assessment of uncertainty, that new walkers err on risky increments due to insufficient walking experience (not poor walking skill), and that infants, like adults, are sensitive to friction forces in the shoe-floor interface and can predict the exaggerated effects of low friction on relatively shallow slopes.
We knew from previous work with 14- and 15-month-olds and adults that low friction dramatically hampers walking skill on slopes. Experiment 1 was designed to establish the extent of the decrement in 18-month-old walkers so that we could determine a new region of uncertainty for Experiment 2. We manipulated friction conditions with slippery shoes rather than a slippery walking surface to make the situation more ambiguous: Infants can see the walking surface, but they cannot see the soles of their shoes; effects of friction must be perceived based on the feeling of slip at the shoe-floor interface (Joh, et al., 2007). Thus, the apparatus was covered with a familiar carpeted surface and infants wore slippery Teflon-soled shoes. To provide a comparison with the slippery shoes and to replicate previous results with this age group (Adolph, 1997; Tamis-LeMonda, et al., 2008), we tested infants walking barefoot and in crepe-soled shoes. Similar walking thresholds in the barefoot and crepe-soled shoe conditions would also rule out the alternative explanation that decrements in the Teflon-soled shoes were due to wearing new (possibly uncomfortable or ill-fitting) shoes rather than due to low friction.
Families were recruited from the greater metropolitan area through visits to the maternity ward of local hospitals and purchased mailing lists. All infants were healthy and born at term. Most families were White, middle-class, and educated. Parents received framed photographs of their infants to commemorate their visit.
Seventy-eight infants participated in one of three walking conditions: Teflon-soled shoes (13 boys, 9 girls), barefoot (18 boys, 22 girls), and crepe-soled shoes (8 boys, 8 girls). Due to practical constraints of recruiting participants, infants wearing Teflon-soled shoes were slightly older (M = 18.82 months, SD = 0.41) than infants walking barefoot (M = 18.04 months, SD = 0.27) and infants wearing crepe-soled shoes (M = 18.07 months, SD = 0.16); F(2,60) = 53.13, p < .001. Post hoc comparisons confirmed these age differences (p < .001). Infants in the barefoot and crepe-soled groups were of similar ages (p > .05). Parents reported infants’ walking experience (first day infants walked at least 10 feet without stopping or falling) in the context of a structured interview, using calendars and baby books to augment their memories. As a consequence of being a few weeks older, infants wearing Teflon-soled shoes had slightly more walking experience (M = 6.54 months, SD = 1.63) than infants wearing rubber-soled shoes (M = 5.08 months, SD = 1.45); F(2,60) = 3.77, p < .05. Post hoc comparisons confirmed this difference in walking experience (p < .05). Walking experience in the barefoot group (M = 5.89 months, SD = 1.69) did not differ from either of the other groups. We considered the slight age and experience advantage of infants in the Teflon group to be acceptable because any impairment due to the Teflon-soled shoes would be conservatively biased—that is, if anything, slightly older and more experienced infants would be expected to show smaller impairments due to the slippery shoes. Seventeen (Teflon = 4, barefoot = 10, crepe = 3) additional infants did not contribute data due to fussiness (N = 12) or experimenter error (N = 5).
We tested infants’ walking skill on a wooden walkway with an adjustable slope (Figure 2). A sloping ramp (86 cm wide × 91 cm long) was connected with piano hinges to flat starting (86 cm wide × 182 cm long) and landing platforms (86 cm wide × 91 cm long) to form a continuous walking surface. An electric motor, operated with a push-button remote, adjusted the slope of the ramp from 0° to 90° in 2° increments by raising and lowering the height of the landing platform from 116 cm to 25 cm. The height of the starting platform remained fixed at 116 cm. A protractor on one side of the walkway indicated the degree of slope. Nets spanned the sides of the walkway as a safety precaution, and the surface was lined with plush carpet. A raised catwalk on one side of the walkway provided the experimenter with easy access to infants in case they fell.
We constructed Teflon- and crepe-soled shoes from flat-soled, commercial infant sneakers (infant sizes 3 to 9). We attached pieces of slippery Teflon material (1.75 mm thick, 20 g) or standard shoe rubber (2 mm thick, 16 g) to the toe and heel sections of the shoe sole, allowing the shoe to flex around the arch when infants bent their feet. All of the shoes looked and felt like typical infants’ sneakers. It was only apparent to infants whether the shoe/carpet interface was slippery after they began walking.
We used a psychophysical procedure developed in earlier work to estimate each infant’s ability to walk down slopes (Adolph, 1997; Adolph, et al., 2008; Tamis-LeMonda, et al., 2008). Trials began with infants standing on the starting platform. Mothers stood at the end of the landing platform and encouraged infants to walk down, offering toys and dry cereal as incentives. An experimenter stood on the raised catwalk and followed alongside infants to ensure their safety. An assistant panned a camera from a side view of the walkway. A second camera recorded the degree of slant from the protractor. Both camera views were mixed onto a single video frame. Trials lasted until infants reached the landing platform or 30 s if they avoided descent. Sessions lasted approximately 90 minutes.
Protocols began with 4 consecutive trials at 0° to teach infants the game of crossing the walkway to their caregivers and to acclimate them to the height of the starting platform. Then, beginning with an easy baseline slope of 4°, the experimenter increased or decreased the degree of slope based on the outcome of the previous trial. Trials were coded online as a success (walked safely), failure (tried to walk but fell), or refusal to walk (slid down or avoided going). After each successful trial, the experimenter increased the slope by 6°. After failures, the experimenter presented infants with the baseline 4° slope to renew their motivation to walk, and then re-entered the protocol with a slope 4° shallower than the previous success. After refusals, infants received the baseline slope and then trials were repeated, so that thresholds were determined based only on the ratio of successes to failures. The rule of “plus 6°, minus 4°” continued until meeting a ≥ 0.67 stopping criterion for estimating the walking threshold: the steepest slope at which the ratio of successes to failures was at least 0.67 and < .67 at the next 2°, 4°, and 6° increments. Fewer trials were required for shallower thresholds and more trials were required for steeper thresholds and protocols with frequent refusals to walk. The total number of trials ranged from 17 to 63, M = 33.14.
We obtained useable protocols from every infant in the Teflon-sole shoe group— presumably because the slippery shoes caused them to fall as slopes became challenging. However, on 12 protocols in the barefoot group and 3 protocols in the crepe-sole shoe group, infants never fell; these protocols were not analyzed because the walking thresholds would be underestimated. Infants who never fell did not differ from infants who did exhibit failures in terms of their age or walking experience. Thus, the final sample consisted of 63 infants, 22 wearing Teflon-soled shoes, 28 walking barefoot, and 13 wearing crepe-soled shoes.
A primary coder rescored trials from video to verify the online scoring. A computerized video coding system, OpenSHAPA (openSHAPA.org), allowed frame-by-frame viewing of the video (Sanderson, et al., 1994). Trials were rescored as success, failure, and refusal as described above. A second coder scored 25% of the trials to verify inter-rater reliability. Inter-rater agreement was 96.7% (κ = .95, p < .001). Disagreements were resolved through discussion. Only threshold estimates derived from video coding were used in analyses.
Teflon-soled shoes grossly impaired infants’ ability to walk down slopes. Despite their slight advantage in terms of age and experience, infants in the Teflon group managed only M = 8.91° slopes (SD = 2.52). Average walking thresholds in the barefoot (M = 23.64°, SD = 5.94) and crepe-soled shoe groups (M = 24.00°, SD = 4.00) were almost three times as steep; F(2,60) = 74.10; p < .01. As shown in Figure 3, while wearing the slippery shoes, the distribution of walking thresholds was dramatically displaced to a shallower range of slopes. In fact, only data from 1 infant in the barefoot group overlapped the distribution of thresholds in the Teflon-soled shoe group. Post hoc comparisons confirmed higher walking thresholds for the barefoot and crepe-soled groups as compared to the Teflon-soled group (p < .001). The similarity in average thresholds between barefoot and crepe-soled shoes suggests that the decrement in the Teflon-sole shoe group resulted from the slippery soles, not the novelty or potential discomfort of wearing new shoes. Moreover, adults in previous work showed similar effects of footgear on walking skill while wearing adult-sized versions of the Teflon- and crepe-soled shoes (Joh, et al., 2007), and 14-month-olds walking in nylon tights over slippery vinyl surfaces showed similar decrements in walking skill due to low friction conditions (Adolph, et al., in press).
Average thresholds were not the only outcome affected by the Teflon-soled shoes: Footgear also affected variability around the group means. As in previous research with 18-month-olds (Tamis-LeMonda, et al., 2008), infants showed a wide range of walking thresholds while walking barefoot (12° to 40°). In contrast, as illustrated in Figure 3, infants showed a compressed range of thresholds while walking in Teflon- (4° to 12°) and crepe-soled shoes (18° to 30°). The narrowed range in the Teflon group is easily explained by the difficulty of walking down slopes in the slippery shoes. Even small movements that threw infants slightly off balance caused them to fall. Similarly, 14-month-olds in previous work showed a compressed range of walking thresholds in low-friction (0° to 6°) compared with high-friction conditions (6° to 26°) (Adolph, et al., in press). But why the compressed range in the crepe-shoe condition? Possibly, infants did not display thresholds in the lower end of the barefoot range because the crepe soles gave the less skilled walkers more traction than barefoot, and infants did not display thresholds in the higher end of the barefoot range because the shoes precluded the more skilled walkers from gaining additional traction by gripping the surface of the slope with their toes (McGraw, 1935).
In previous work, experienced 18-month-old walkers used social information selectively only at the region of uncertainty around their walking thresholds (Tamis-LeMonda, et al., 2008). But those infants were walking barefoot and their walking thresholds were in a mid-range of slopes (20° to 30°). The question here was what would happen when we altered infants’ walking skill with the Teflon-soled shoes: Would 18-month-olds be transformed into novice walkers and behave like the 12-month-olds in previous work (Adolph, et al., 2008)? Would they act oblivious to the feeling of slip, as we had observed in 14-month-olds previously (Adolph, et al., in press), and simply respond to the degree of slant? Or would they update their assessment of uncertain risk when walking down slopes wearing slippery shoes?
As in previous research (Adolph, et al., 2008; Tamis-LeMonda, et al., 2008), the primary outcome measure was whether infants attempted to walk. In addition, we scored infants’ responses on trials where they refused to walk to determine whether mothers’ discouraging messages caused infants to avoid descent by remaining on the starting platform or whether infants used alternative sliding positions to descend. We also scored infants’ spontaneous exploratory activity—latency and touching—to describe how they generated perceptual information for potential risk, and their social expressions—positive and negative facial expressions and vocalizations—to determine whether infants’ social expressions mirrored mothers’ social signals and/or accompanied risk avoidance.
Twenty-four 18-month-olds (M = 18.44 months, SD = 0.44; 12 boys and 12 girls) and their mothers were recruited as in Experiment 1. Walking experience (M = 5.93 months, SD = 1.77) was similar to infants in Experiment 1. Infants were relatively experienced walkers. They were primarily White and from middle-class homes. Six additional infants did not complete testing due to fussiness (N = 5) or experimenter error (N = 1).
Infants wore the same slippery, Teflon-soled shoes and walked over the same carpeted, sloping walkway used in Experiment 1. As before, an experimenter followed alongside infants to ensure their safety. Sessions lasted approximately 90 minutes. Because the range of thresholds was so compressed in the Teflon-soled group in Experiment 1, for Experiment 2, it was not necessary to estimate thresholds on an individualized basis before testing effects of social signals (as in Adolph, et al., 2008; Tamis-LeMonda, et al., 2008). Instead, we could safely assume that infants would only be able to manage very shallow slopes. Thus, testing consisted of 4 blocks of 11 trials distributed in 5° increments from 0° to 50° in 1 of 4 random trial orders. The shallowest slopes (0°, 5°, 10°, and 15°) included 100% of walking thresholds from infants in the Teflon group in Experiment 1—presumably uncertain for infants in Experiment 2. The mid-range of slopes (20°, 25°, 30°, and 35°) included 93% of thresholds from infants in the barefoot and crepe groups in Experiment 1—safe under normal footgear conditions, but impossible while wearing the slippery shoes. The steep range of slopes (40°, 45°, and 50°) were impossible for all but one barefoot child in Experiment 1—that is, impossible in any footgear.
Trial blocks alternated between encourage and discourage conditions. Condition order and trial presentation orders were counterbalanced across infants. Trials lasted 30 s or until infants initiated descent, whichever occurred first. An assistant panned a camera from the side of the walkway to record infants’ whole body movements, a second assistant recorded a close-up view of infants’ faces from the bottom of the walkway, and a third camera recorded the degree of slant from the protractor attached to the walkway. A shotgun microphone mounted on the ceiling recorded infants’ vocalizations. The camera views and audio track were mixed onto a single video frame.
Protocols began with 4 consecutive warm-up trials on the flat 0° walkway to ensure that they could walk in the slippery shoes and to acclimate them to the raised walkway. As in previous work (Adolph, et al., 2008; Tamis-LeMonda, et al., 2008), on test trials, mothers were instructed to encourage/discourage their infants to walk down the slopes using words, vocal intonations, facial expressions, gestures, and body language in whatever way felt most natural. In the encourage condition, mothers were instructed to “get their infants to try to walk down the slopes;” they should “treat the situation as if they were encouraging their infants to tackle a new challenge such as taking their first walking steps or jumping into a caregivers’ arms at the swimming pool.” In the discourage condition, mothers were instructed to “prevent their infants from trying to walk down the slopes;” they should “treat the slope as if it were a sheet of ice that would jeopardize infants’ safety.” Mothers were told to disregard the steepness of the slopes when delivering their messages. Previous work verified that mothers’ behaviors varied by social incentive condition but not by slant (Karasik, Tamis-LeMonda, Adolph, & Dimitropoulou, 2008; Tamis-LeMonda, et al., 2008).
Mothers sat on a raised platform adjacent to the landing platform, at infants’ eye level but out of reach (Figure 4). At the beginning of each trial, the experimenter held infants on the starting platform. After infants faced their mothers, an assistant rang a bell to signal mothers to begin encouraging or discouraging. The experimenter released the infant after 2 s and mothers continued to deliver their social message.
A primary coder scored each trial from video using OpenSHAPA. A secondary coder scored 25% of each infant’s data to verify inter-rater reliability. Disagreements between coders were resolved through discussion. Attempts to walk were scored as successes, failures, and refusals as described above. Coders agreed on 96.5% of trials (κ = .94). On trials scored as refusals, coders determined whether infants slid down (sitting, head first prone, or feet first backing) or avoided descent for the duration of the trial (98.5% agreement, κ = .98). Latency to descend was scored from the start of the trial until infants initiated descent, and included the time infants explored slopes by looking and touching. Latency could range from 0 s (immediate decision) to 30 s (maximum trial length). The correlation between coders’ scores was r(184) = .98, p < .001. Infants nearly always used their feet, not their hands, to explore slopes by touching. Thus, tactile exploration included rubbing the feet against the slope, rocking back and forth with feet straddling the brink, or taking tiny steps at the edge of the slope (99.5% agreement, κ = .83). Coders scored negative vocalizations as whining, whimpering, and crying (97.0% agreement, κ = .90) and negative facial expressions as frowns, pouts, and downward curls of the mouth (94.5% agreement, κ = .83). All other vocalizations and facial expressions were scored as positive/neutral (92.0% agreement, κ = .84; and 100% agreement κ = 1.0, respectively). Both negative and positive/neutral social expressions could occur in one trial.
Preliminary analyses showed no effects of presentation order, so data were collapsed across orders for subsequent analyses. Result graphs are shown with all 11 slopes (0° to 50°), but for statistical analyses, we compared infants’ behaviors across shallow (0° to 15°), intermediate (20° to 35°), and steep slopes (40° to 50°), based on the range of thresholds obtained in Experiment 1. Consistent with Experiment 1, success rates were 92% at 0°, 86% at 5°, 32% at 10°, and 0% on steeper slopes. For each outcome measure, data were subjected to a 2 (social message condition: encourage and discourage) × 3 (risk levels: shallow, intermediate, and steep) repeated measures ANOVA. Interactions were followed up with paired t tests between conditions at each risk level (with Bonferroni-adjusted alpha levels of p < .02).
Attempts to walk were more frequent in the encourage condition and decreased linearly with risk from 1.0 at 0° to .17 at 50° (open square symbols in Figure 5A). If we had only tested infants with encouraging social signals (as in Adolph, et al., in press), we would have concluded that infants responded only to slant, not to friction. Fortunately, here we also tested infants with discouraging social signals. As shown by the closed circle symbols in Figure 5A, infants’ responses to discouraging social signals revealed sensitivity to the low friction manipulation: At 0°, when mothers discouraged, attempt rates dropped from 1.0 to .59 and decreased steadily across risk levels. In fact, mothers’ social messages had the largest effect on infants’ attempts at the shallowest slopes, indicating that infants had updated their assessment of uncertainty in accord with the novel situation of walking in slippery shoes and that they used social information accordingly.
As in previous work (Tamis-LeMonda, et al., 2008), infants continued to defer to mothers’ messages at the mid-range of slopes, where risk would normally be uncertain, but where walking was now impossible. But, unlike the 12-month-old novice walkers shown in Figure 1B (Adolph, et al., 2008), 18-month-olds in slippery shoes ignored their mothers’ social signals at the steepest range of slopes, indicating that poor walking skill is not the factor responsible for high error rates in novice walkers. This interpretation is corroborated by similar evidence from 14-month-olds wearing heavy shoulder packs that hindered their walking skill; infants flexibly updated their perception of uncertain risk, treating the same degrees of slope as safe while wearing feather-weight packs but as impossible while wearing lead-weighted packs (Adolph & Avolio, 2000).
The ANOVA confirmed main effects for condition, F(1, 23) = 19.26, p < .001, partial η2 = .46; risk level, F(2, 46) = 42.64, p < .001, partial η2 = .65; and a condition by risk level interaction, F(2, 46) = 13.35, p < .001, partial η2 = .37. Post hoc comparisons confirmed higher attempt rates when mothers encouraged at the shallow and intermediate ranges of slopes (Ms = .83 versus .46 and .42 versus .25 for encouraging and discouraging messages at the shallow and intermediate ranges of slopes, respectively).
A second source of evidence that infants updated their perception of uncertainty and used social information selectively is their high rate of avoidance on shallow slopes. On trials where infants refused to walk, they could slide down (in sitting, backing, or prone positions) or avoid going entirely. Overall, infants avoided descent on 57.2% of refusal trials. The striking finding is the high rate of avoidance on the 0° to 15° slopes when mothers discouraged (M = .45) compared to when she encouraged (M = .09) and the fact that mothers were equally successful at keeping infants on the starting platform across risk levels in the discouraging but not the encouraging condition (Ms = .48 versus .26 and .52 versus .27, for discouraging versus encouraging at the intermediate and steepest slopes, respectively).
Indeed, avoidance rates mirrored attempt rates for both types of social signals. Figure 5B shows that half of the sample (y-axis shows number of infants) avoided descent on at least one trial at every risk level in the discouraging condition. When mothers encouraged infants to walk, infants refused to do so at the steepest increments, but they frequently slid down (65% of refusal trials) rather than avoiding descent, indicating that the encouraging social signals did affect their behavior by prompting them to find alternative means of descent. An ANOVA on the proportion of avoidance trials revealed main effects for condition, F(1, 23) = 18.88, p < .001, η2 = .45; risk level, F(1, 46) = 6.89, p < .01, η2 = .23; and an interaction between condition and risk, F(2, 46) = 3.74, p < .05, η2 = .14. Post hoc comparisons showed higher rates of avoidance when mothers discouraged at every risk level, with the largest disparity at the shallow range of slopes.
As in previous work (e.g., Adolph, 1997; Tamis-LeMonda, et al., 2008), latency to descend increased with risk in the encouraging condition (Figure 5C). But mothers’ social signals had differential effects. Regardless of whether they ultimately decided to walk, slide, or avoid, at every risk level, infants hesitated for longer durations while formulating their decisions during discouragement (M = 17.3 s) compared with encouragement (M = 8.89 s). During the time that they hesitated, infants looked at the slope and at their mothers and occasionally called to their mothers. Most important, as shown in the figure, latency mirrored attempt rates. The largest disparity between encouraging and discouraging social signals was at the shallowest range of slopes. Remarkably, discouraging social signals kept infants poised in indecision on the starting platform at 0° (14.67 s) for nearly the same duration as at 50° (18.67 s). As a comparison, when mothers encouraged, infants hesitated for only 1.76 s at 0° before running down; at 50°, they hesitated for 11.74 s and refused to walk. An ANOVA confirmed main effects for condition, F(1, 23) = 30.61, p < .001, η2 = .57; risk level, F(1, 46) = 15.35, p < .001, η2 = .40; and an interaction between the two factors, F(2, 46) = 4.38, p < .05, η2 = .16. Post hoc comparisons showed differences between social message conditions at every risk level, with the largest difference at the shallowest slopes.
In contrast to previous work with experienced walkers (Adolph, 1997; Adolph & Avolio, 2000; Tamis-LeMonda, et al., 2008), touching was infrequent at every risk level, regardless of social message (Figure 5D). The ANOVA showed no effects for risk level, F(2, 46) = 1.87, p > .05; social message, F(1, 23) = 0.25, p > .05; or interaction F(2, 46) = 0.86, p > .05. Overall, infants touched on only 12.8% of trials, and 4 infants never touched. Possibly, touching was constrained by the difficulty of keeping balance in the slippery shoes while poking one foot out or rocking at the brink. Fourteen- and 15-month-old infants also showed depressed levels of tactile exploration on slippery slopes compared with high friction conditions (Adolph, et al., in press).
As shown in Figure 6A–B, across conditions and risk levels, infants consistently displayed positive-neutral facial expressions (M = .99) and rarely displayed negative facial expressions (M = .11). Six infants never produced a frown, pout, or downward mouth curl, and negative facial expressions only exceeded .50 at any increment for 2 infants. These 2 were not indiscriminate “cry babies” because they also displayed positive-neutral facial expressions at the same increments. The ANOVAs showed no effects of social condition, F(1, 22) = 3.01, p > .05; risk level, F(2, 44) = 0.38, p > .05; or their interaction, F(2, 44) = 0.38, p > .05, on infants’ positive-neutral facial expressions or negative facial expressions, Fs = 0.18, 0.57, 0.14, ps > .05, respectively. The results corroborate previous findings that infants produce positive-neutral—not negative—facial expressions while refusing risky increments, even when their mothers display negative social signals (Adolph, et al., 2008; Sorce, et al., 1985; Tamis-LeMonda, et al., 2008).
Infants’ vocalizations were similarly lacking in negativity (compare Figures 6C–D). Overall, infants vocalized on 40.3% of trials. Of these, 29.2% contained only positive-neutral vocalizations, 4.5% contained only negative vocalizations, and 6.5% contained both. Nine infants never produced a negative vocalization.
As shown in Figure 6C, across risk levels, positive-neutral vocalizations were more prevalent when mothers discouraged (M = .41) than when they encouraged (M = .27). At first blush, the result seems counter-intuitive, but it makes sense when considering vocalizations and latency together. Given the low rate of negative vocalizations overall and the fact that mothers kept infants on the starting platform for longer periods in the discourage condition, infants had more time to vocalize in the discourage condition and what came out of their mouths was positive-neutral babbles rather than negative whimpers. The ANOVA revealed only main effects for condition, F(1, 23) = 7.37, p < .05, η2 = .25; and risk level, F(1, 46) = 10.99, p < .001, η2 = .33. As shown in Figure 6D, negative vocalizations did not depend on mothers’ social signals; the ANOVA revealed only a main effect for risk level, F(1, 46) = 3.72, p < .05, η2 = .15.
Traditionally, studies of social-cognitive development and locomotor development are worlds apart, using different methods to address different theoretical issues. The current study brought these two worlds together. We used a locomotor task—walking down slopes in slippery Teflon-soled shoes—to ask whether infants’ selective use of social information is dynamically tuned to the current constraints on action. At the same time, we manipulated the social information available to infants—mothers encouraged and discouraged their infants to descend— to disentangle walking experience from walking skill and to examine infants’ sensitivity to friction underfoot.
Experiment 1 showed that Teflon-soled shoes severely impair walking skill in experienced 18-month-old walkers: Average walking thresholds decreased from 24° while barefoot or wearing standard crepe-soled shoes to 9° in the Teflon-soled shoes. Thus, the normal region of uncertain risk (20° to 30°) was displaced to a shallower region of slopes in slippery shoes (< 15°). In Experiment 2, 18-month-olds updated their assessment of uncertainty to take their diminished walking skill into account, and they used social information selectively at the shallow region of slopes. These results indicate that adaptive responding depends on walking experience, not walking skill, and that infants are sensitive to the feeling of slip beneath their feet.
Locomotor tasks provide new ways to operationalize and conceptualize uncertainty. The concept of uncertainty is central to studies of social referencing because infants should only seek social information or defer to unsolicited advice when self-generated perceptual information is insufficient (Baldwin & Moses, 1996). When risk is not uncertain (i.e., the situation is clearly safe or impossible), then infants should instead rely on perceptual information from their own exploratory activity. However, in traditional paradigms, the selection of uncertain toys and people relies on researchers’ subjective intuitions about what stimuli will seem ambiguous to infants. Moreover, a single bizarre toy or strange person cannot reveal dynamic or precise attunement to the level of uncertainty.
By contrast, uncertainty can be objectively quantified in locomotor tasks across a graded array of risk levels and infants can be tested in dozens of trials (Adolph, 1997, 2000; Adolph & Avolio, 2000; Adolph, et al., 2008; Tamis-LeMonda, et al., 2008). In other words, uncertainty can be operationally defined in terms of the probability of walking (or crawling, reaching, etc.) successfully under varying conditions. As demonstrated in the current study, any particular increment of risk (e.g., a single degree of slope) cannot provide the complete picture of infants’ assessment of uncertainty. The whole array is needed to reveal whether infants’ assessment of risk matches the underlying probabilities in terms of motor skill, and whether infants use social information only at uncertain increments.
Perhaps most important, the current study highlights the dynamic nature of objective uncertainty. Simple manipulations such as varying the slope of the ground surface or the friction at the shoe-floor interface can change the probability of walking successfully. More generally, possibilities for action change with every variation in infants’ abilities and in features of the environment. Infants’ effective use of social information requires them to track these changes in affordances to guide actions adaptively.
A remarkable finding was infants’ cautious behavior at 0° slopes when mothers discouraged. Every infant could walk over the flat walkway—they did so multiple times in the warm-up trials. Nonetheless, mothers’ discouragement elicited avoidance in half of the infants and increased latency and sliding strategies in the other infants. A similar pattern held across the entire range of shallow slopes: Discouraging social signals led to more cautious and conservative responses.
Two mechanisms may underlie the power of discouragement for making infants think twice at the shallowest increments. Possibly, discouragement focused infants’ attention on mothers’ face and voice, rather than on the slope. Indeed, some mothers used sharp, abrupt warnings (“Lily, No! Stop!) and stopping guard hand gestures when infants moved forward, followed by calm, soothing, elongated utterances (“Goood, goood, staaay”) when infants stopped moving (see Karasik, et al., 2008). The overall effect was reminiscent of a dog owner executing the sit-stay command to a pet. A second possibility is that discouragement—coupled with the visual cues for the flat surface or shallow slant—led to increased information gathering by the infants. A consequence of spending a longer period on the starting platform is that infants had more opportunities to gather perceptual information. Indeed, mothers frequently referred to the slope with words and gestures, eliciting looks by infants toward the obstacle. Thus, regardless of whether infants ultimately deferred to mothers’ discouraging advice, mothers may have succeeded in getting infants to look before they leaped.
A widespread interpretation of infants’ avoidance of impossible obstacles (in particular, on the deep side of the visual cliff) is that infants are afraid of falling and that mothers’ discouraging messages heighten their natural fear response (Bertenthal & Campos, 1984; Campos, et al., 2000; Saarni, Mumme, & Campos, 1998; Sorce, et al., 1985). Several lines of evidence, however, argue against a fear-based interpretation of infants’ behavioral reactions. In the Sorce et al., (1985) study, for example, infants did not exhibit outward signs of fear. Their affect changed from “happy” to “neutral” when mothers posed joyful versus fearful/angry facial expressions. Similarly, when mothers provided encouraging social messages on the deep side of the visual cliff, neither crossing nor avoidant 9- to 12-month-old infants displayed fearful expressions or accelerated heart rate (Rader, Bausano, & Richards, 1980; Richards & Rader, 1981, 1983). In previous work, 12- and 18-month-olds displayed positive-neutral facial expressions and vocalizations on impossibly steep slopes (Adolph, et al., 2008; Tamis-LeMonda, et al., 2008). In other words, outward signs of fear do not accompany adaptive motor decisions and negative social expressions do not differ between infants who respond adaptively and those who do not. Saarni et al., (2006) argued that infants’ smiles and other outward signs of positive affect do not refute a fear-based account because the avoidant behavior itself indicates fear. However, given the circularity of their argument, it is prudent to rely on independent measures of affect beyond avoidance itself.
In the current study, we found no evidence that mothers’ discouraging message or the sight of a precipice instigated fear or any other negative expression. Infants’ facial affect did not differ between the encouraging and discouraging conditions; their faces were positive not negative. Positive vocalizations were higher in the discouraging condition and negative vocalizations were low in both conditions. Likewise, we found no evidence that steep slopes instigated negative social expressions or that refusals to walk were accompanied by fear. In fact, positive vocalizations increased with increasing risk, and negative vocalizations were consistently low across risk levels.
Infants’ selective use of social information at the shallowest increments sheds light on two issues regarding locomotor development. The first issue concerns the effects of locomotor experience (how many days since infants began walking) versus locomotor skill (infants’ walking thresholds on slopes). Normally, the two factors are confounded such that novice walkers have little experience in their new upright posture and they also display poor levels of locomotor skill. By dressing infants in the Teflon-soled shoes, we transformed experienced, proficient 18-month-old walkers into tippy, poorly skilled walkers, without changing their upright posture or eliminating their months of walking experience.
The 18-month-olds did not respond like the novice 12-month-old walkers in previous work by using social information selectively only at the steepest slopes, and even then with persistently high attempts to walk under impossible conditions (see Figure 1B). In fact, mothers’ social signals had no effect on 18-month-olds’ decisions between 30° and 50°, where infants did not budge regardless of mothers’ social message; rather, infants showed the largest effects at the shallowest slopes. The findings lend support to the proposal that it is lack of experience not skill that is responsible for rash responding to potential risk in novice walkers.
A second locomotor issue that is addressed by infants’ selective use of social information is their sensitivity to friction underfoot. In previous work where mothers only encouraged infants to descend, 14- and 15-month-olds responded only to slant, not to friction (Adolph, et al., in press). They attempted to walk down the same degrees of slope regardless of whether the condition was high or low friction, and they fell on the low friction surface on trial after trial. The current study showed that older, more experienced walking infants do display sensitivity to the slip between their feet and the ground surface, as evidenced by the disparity between attempt rates at the shallow slopes under conditions of mothers’ encouragement and discouragement. Nonetheless, at the 10° to 20° slopes where the probability of success was negligible, attempt rates were relatively high (.33 to .77), even in the discouraging condition. Thus, although infants displayed sensitivity to friction that was not revealed in past research that did not manipulate social information, infants’ responses were far from adaptive. Apparently, visual information for slant is still a more potent source of information for guiding infant locomotion than tactile information for friction.
The traditional approach to motor development is overly simplistic: Clearly social information plays a role in guiding locomotion. But, the importance of social information in infants’ decisions about action should not be overstated. When infants are presented with conflicting information—social signals from their caregivers versus perceptual information generated by their own exploratory activity—the latter generally wins the day. In no study thus far, have infants been found to defer to social information across the board. Such a conclusion is probably good news. In everyday situations, parents cannot be so vigilant that they can protect infants from every potential danger. With the advent of independent locomotion comes increasing autonomy. Although mothers’ advice can often be useful, especially under conditions of uncertainty, infants must eventually learn to navigate the world on their own.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.