Poster workshop
Chairs: Amanda L. Woodward and Beate Sodian
Discussant: Beate Sodian
Infant representations of action likely lay the groundwork for language,intentional understanding, social competence, and problem solving, amongother critical abilities. Currently, researchers are investigatinginfants' action representations from many different vantage points. Theseresearchers start from several distinct theoretical backgrounds and bringto bear a range of empirical methods. This poster workshop aims atbringing together several diverse lines of research, with the goals of bothsurveying the current empirical landscape and promoting contact betweendifferent vantage points on the phenomenon. To this end, we have broughttogether 11 researchers who span a range of questions and methods instudying infants' action representations. The posters in this workshopspeak to four interrelated sets of issues.One set of issues concerns when and how infants begin to conceive of othersas intentional beings, acting in a purposeful way towards objects and otherpeople. A traditional approach to this question focuses on infants'understanding of social-communicative intentional actions such as gaze andpointing. Traditionally, this understanding has been tapped byinvestigating infants' developing competencies in triadic interaction. Recently, several laboratories have begun to use visual habituationmeasures as an additional source of evidence about infants' knowledge ofthese actions. Several posters in the workshop take one or both of theseapproaches in studying normally developing infants as well as children withautism. In addition, several posters consider other potential hallmarks ofintentional understanding including knowledge about the object-directednessof actions and knowledge about action as a means to an end.The second set of issues concerns infants' knowledge about the structure offamiliar intentional actions such as pointing, gaze, emotional expressions,and giving. Several posters describe work that taps infants'representations of these actions, asking whether infants represent theelements of the action that are critical from an adult standpoint. Thesestudies provide evidence on infants' ability to learn about particularactions, the structure of infants' action categories, and the prelinguisticknowledge that provides a basis for both language learning and furtherdevelopments in intentional understanding.The third set of issues concerns infants' ability to encode novel actionsin a way that will facilitate further analysis. Are infants able to parsethe ongoing stream of events into units that correspond to intentionalunits from an adult standpoint? What are the perceptual cues that underliethis parsing ability? Posters in the workshop provide evidence on thesequestions, as well as evidence on infants' ability to use means-endrelations to interpret novel actions. The fourth set of issues concerns the relation between infants' actionrepresentations and their category knowledge about agents and objects. Actions relate agents to objects. How does infants' understanding ofaction provide a basis for reasoning about agents and the objects they acton? First, does the presence of actions such as social reciprocity andgiving lead infants to infer that a novel object is an agent? And second,does infants' knowledge about objects include an understanding of the waysin which agents typically act on them? Posters in the workshop addressthese questions.In summary, the present posters address a wide range of important issues inthe fast developing field of infant action knowledge. The poster workshopoffers the opportunity for researchers from across this field to sharefindings and perspective on the issues, as well as an opportunity forothers to survey the findings and debates in the field.
Details of individual items:
poster
When do young children understand that others act in purposeful orintentional ways towards objects? Recently, researchers have found that24-month-old children use a person's head turns and points to determinewhich object she desires (Lee et al., 1998). Also, it has been shown that18-month-old infants, based upon a person's gaze, gesture and vocalization,are able to predict which of two objects a person will reach for or label(Baldwin et al.., 1996;Poulin-Dubois, 1999). These studies suggest thatinfant's understanding of a person's intentions/desires depends much on thebehavioral cues available to them. We first report data showing that18-month-olds can predict the object a person desires when both gaze andgesture are directed toward the target object. In another study, weexamine the relative importance of behavioural cues to intentions (e.g.,gaze, gesture, emotional expression, and vocalization) in 14- and18-month-old infants in a toy-request task. In each of eight toy-requesttrials, infants watch a female experimenter express interest in one of thetwo toys (e.g., looking at and gesturing towards) that are out-of-reach ofthe experimenter. The experimenter then holds out her hand and asks theinfant to give her one object. Sixty infants (thirty in each age group) arerandomly assigned to one of two conditions. In one condition, the targettoy remains in the same location throughout the experiment. In the secondcondition, the location of the target toy is switched after the display ofthe behavioural cues. In each condition, the experimenter displays fourclusters of behavioural cues (i.e, gaze and vocalization; gaze and gesture;gaze and emotion; and gaze, vocalization, emotion, and gesture) towards thetarget toy. We predict that 18 month-old infants will be more likely than14-month-old infants to 1) use the more impoverished cues (e.g., gaze andgesture) to identify the desired object and 2) link the experimentersbehavioural cues to the desired object and not to a specific location.
poster
Intention is one of the earliest mental states to be understood by infants. In this study, we investigate understanding of others' intentions intypically-developing infants and we compare this understanding to that of aspecial population of children who are known to have difficulty with theunderstanding of others' mental states: children with autism. Many studieshave shown that children with autism have difficulty understanding others'thoughts and beliefs. However, far less research has been conducted onwhat these children understand about 'simpler' mental states such asintentions. In one previous study, Phillips, Baron-Cohen, and Rutter(1992) investigated what they termed goal detection in typically-developinginfants, children with autism, and children with other developmentaldelays. In their tests, an adult performed two ambiguous actions (blockingthe child's play and withdrawing a toy in a teasing fashion) and oneunambiguous action (giving the child a toy). They coded children's gaze tothe adult's face during the 5 seconds following the adult's actions. Theyfound that typically-developing infants and most children withdevelopmental delays looked to the adult's eyes immediately after theadult's ambiguous actions but not her unambiguous action. Children withautism, in contrast, rarely looked to the adult's eyes in any of thesesituations. Since then, Charman et al. (1997) have failed to replicatethese results, finding no between-group differences in gaze to the adult. Because of these mixed findings, and because it is difficult to know whychildren looked to the adult in this situation (e.g., goal detection versusimperative requesting versus social referencing), it is unclear whatchildren understood about the adult's actions. In the current study, wereplicated the procedure of Phillips et al. (1992) and also added severalother experimental conditions in order to address alternate explanations ofwhat children's looks to the adult meant. Using the same general procedureof coding children's responses to adults' actions, we added imperative,social referencing, and joint attention conditions. We also testedchildren's understanding of others' intentions more directly by comparingchildren's responses to intentional and accidental actions. We found veryfew differences in looking patterns between children with autism andchildren with developmental delays. However, there was some evidence thatthese two groups of children were looking to the adult for differentreasons. We compare the results for these children to those fortypically-developing infants and discuss the implications for theories ofautism and typical development.
poster
Many studies have explored infants' understanding of the point gesture as aspotlight to a referent. These studies have measured infants' propensityto follow points produced by social partners. However, little is knownabout infants' understanding of the relation between the person who pointsand the referent. Adults understand pointing not only as a signal to turnin a particular direction, but also as an action that relates a person tosome object. When do infants understand pointing in this way? The currentstudies investigated this question using the visual habituation paradigm. In the first study, 9- and 12-month-old infants were tested. Infants sawan actor point to one of two toys during habituation, and then saw testevents in which either the actor moved in the same way to point at a newobject (new referent trials), or she moved her arm through a new path topoint at the same object as during habituation (old referent trials). Newreferent trials altered the relation between actor and referent whilepreserving the surface features of the habituation event, whereas the oldreferent trials altered the surface features of the habituation event whilepreserving the relation between actor and referent. Twelve-month-oldslooked longer on new referent trials than old referent trials, indicatingthat they had attended selectively to the relation between the actor andthe referent. In contrast, 9-month-olds looked equally at the two testevents, suggesting that they did not see the relation between the actor andreferent as more critical to the event than its surface features. Followup analyses suggested a relation between infants' point production duringthe study and their patterns of looking on test trials. Those 9-month-oldinfants who produced many points looked longer on new referent trials thanthose who produced few points. To further investigate the development of infants' understanding ofpointing, a second study was conducted, testing infants at three ages: 7, 81/2, and 10 months. To follow up on the prior finding of a relationbetween the infants' experience producing points and the experimentaleffect, information was also gathered regarding infants' point productionat home. The habituation task revealed a developmental pattern similar tothat found in Study 1. Seven-month-old infants did not understand thepointing gesture as implying a relation between actor and referent: theylooked equally on new referent and old referent trials. Eight-and-a-half-month-olds looked marginally longer on new referenttrials. Ten-month-olds looked reliably longer on new referent trials. Study 2 provided additional evidence for the relation between infants' ownpointing and their understanding of pointing in others. Infants whopointed at objects at home tended to look longer on new referent trialsthan infants who did not point. Taken together, the findings of the twostudies indicate that between 8 and 10 months infants are coming tounderstand pointing as a behavior that relates an actor to a referent, andthat this insight may be tied to infants' own experience producing thepoint gesture.
poster
This study explored 14-month-olds' ability to categorize an object asmentalistic on the basis of its actions alone, regardless of itsappearance. Several converging measures of putatively mentalisticcategorization were taken from infants' reactions to a novel object. Thirty 14-month-old infants were introduced to a small purple hand-sized,remote-controlled, blob-like entity in a goal-re-enactment paradigmmodelled after Meltzoff (1995). In a warm-up, the experimenter, entity, andinfant took turns 'passing' a toy block back and forth. Infants then sawone of three events; (i) the entity successfully knocked the block into acup (the modelled goal condition), (ii) tried unsuccessfully to knock itinto the cup (the inferred goal condition), or (iii) danced around theblock in a circle without regard for the cup (the baseline controlcondition). After each observation, the infant was given the block to playwith for 20 seconds. Several measures of the infants' reactions were made,including the frequency with which they (i) placed the block in the cup,(ii) looked toward the cup, (iii) alternated their gaze between the entityand the block, or (iv) perserverated with the warm-up game by passing theblock back to the entity. The number of times overall that the infantwithdrew from the entity was also measured.Overall, 83% of the infants withdrew from the entity at least once duringthe procedure--in the same manner that infants of this age withdraw fromstrange humans. Ninety percent of the infants alternated their own gazebetween the entity and the toy block at least once--in the same manner thatinfants check the gaze of humans in novel situations.Unfortunately, many ofthe infants perserverated on the warm-up task during the test trials,making one of the measures of primary interest (the rate of goalre-enactment) difficult to measure. However, when those infants (17 in all)were removed, the resulting percentages of infants placing the block intothe cup were 88% in the modelled goal condition, 8% in the baselinecondition, and 50% in the inferred goal condition. The inferred conditiondid not differ from the modelled condition but did differ marginally fromthe baseline control (p < .08). This difference was further supported bythe direction of gaze of all thirty infants during the test trials. Infantsin the baseline condition looked toward the goal (i.e., the cup) only 0.7times on average, while infants in the other two conditions (whoseattention had been drawn toward the cup) looked far more often, 3.75 timeson average. Furthermore, among those later infants, infants in the inferredgoal condition looked nearly twice as often as did infants in the modelledgoal condition, 4.6 vs. 2.9, F(1,18) 4.54, p < .05, suggesting that theymay have recognized the role of the cup even as they gave the block back tothe entity.These results suggest that infants may be able to categorize novel entitiesas mentalistic agents on the basis of their actions alone, regardless oftheir lack of perceptual similarity to known mentalistic objects such aspeople.
poster
Children's play activities involving toy-objects have traditionally beenstudied by developmental psychologists interested in the emergence ofsymbolic representations and conceptual categories during the second yearof life. Recently, free play with toys has also been analyzed incategorization studies with preverbal children. Here, the main focus is howlong infants examine a given object. Examination time is the time duringwhich infants concentrate on and inspect a given item. In a typicalobject-examination task, infants play freely with different-lookingminiature toy-models representing exemplars from the same category beforethey get an out-of-category item during the test-phase. If examination timeincreases from the last habituation trial to the test-trial, this is takenas evidence for infant's awareness of differences between both categories.Using this task, Mandler and McDonough (1993) found a global-to-basic levelshift in infants as young as 7-11 months of age. In the present study, wetested whether infants aged 9 months (N80) and 11 months (N80)distinguish among the global categories of animals, plants, and artifacts.In a modified object-examination task, infants were habituated to 10different exemplars from one category before an out-of-category item waspresented during trial 11. If examination time decreased from the first tothe second half of habituation trials, this was taken as evidence forhabituation. If examination time increased significantly from trial 10 totrial 11, this was taken as evidence for category discrimination. Analysisof variance revealed that infants habituated to animals only, anddishabituated to plants as well as to artifacts as test-stimuli. Plants andartifacts neither elicited a habituation response, nor did infantsdiscriminate them from each other, as indicated by the lack of adishabituation response during the test-phase. This does not imply,however, that preverbal children know nothing about exemplars from thesetwo categories: In addition to infant's responses regarding examinationtime, we also analyzed the frequency of functionally appropriate actionsinvolving the presented toy-artifacts. A categorization schema for theseactions was developed in advance. Each trial was counted only if twoindependent coders identified a given action as functionally appropriatefor the given artifact (range: 0 to 10 per child). Among those 40 infantswho were habituated to artifacts, the mean number of activities showingfunctional knowledge at the basic level (e.g., combing hair with a brush,holding a telephone to your ear) was significantly higher for 11-month-olds(M2.15) than for 9-month-olds (M0.85), t(38)3.0, p<.01. This suggeststhat functional play revealing basic-level knowledge within the artifactdomain can be observed much earlier than traditionally assumed.Furthermore, it seems to undergo substantial developmental changes beforethe end of the first year of life. The application of such knowledge mayexplain the lack of categorization responses in the presented global-levelobject-examination task. For future studies using this method, it may thusbe important to analyze not only changes in mean examination time but alsoplay activities with the presented toy-stimuli.
poster
This research investigates some of the earliest ways in which infants mayreason about the behavior of adults, specifically, whether or not they areable to utilize information about affect and perception to predict action. Four studies were conducted addressing this question. Study 1 tested thehypothesis that 12-month-olds reason with the principle that people arelikely to reach for objects which they like or desire and thus regard withpositive affect. We used an infant controlled habituation method closelyparalleling that used in many studies of physical reasoning. Infants werehabituated to an event in which an adult looked to one of the two stuffedanimals with an expression of interest and joy. The adult was subsequentlyshown holding that kitty. The infants were then shown two types of testtrials,an event which, like habituation, was consistent with the principlesthat people act where they look and express positive affect. Infants inthis study looked longer at the inconsistent event, providing evidence that12-month-olds expect an adult to pick up the object regarded. Furtherstudies replicate this result and attempt to control for exposure topick-up during habituation. In addition we have found that 14-month-oldsexpect a person to pick up the object regarded, even without habituation. Together these studies provide evidence that as early as twelve-months,infants have expectations about intentional action and that thisunderstanding is enriched between 12 and 14 months.
poster
One fundamental question concerning early socio-cognitive development ishow infants come to understand the goals and intentions underlying humanbehavior. This is in part complex because behavior typically proceeds in acontinuous stream with no transparent prepackaging of distinct actions. Toderive meaning from the behavior stream infants mustfirst organize it into relevant units for analysis. We recently utilizedhabituation/dishabituation and cross-modal matching methodologies to testwhetherinfants are able to parse continuously flowing action into units thatcoincide with the initiation and completion of the actor's intentions. Using a variant of the habituation/dishabituation paradigm, we gave 10- to11- month-old infants the opportunity to view digitized video sequences ofcontinuous action (e.g., a woman reaching for, grasping, and placinga towel on a rack). After familiarizing to this action sequence, infantsviewed two test versions in which still-frame pauses were introduced toeither (a) coincide with completion points of intentional action or (b)interrupt ongoing intentional action. Infants looked longer at the versionin which the pause interrupted the pursuit of intentions than at the one inwhich the pause coincided with the completion of intentions. Thesefindings indicate that infants spontaneously process action in terms ofunits bounded by the initiation and completion of intentions; disruptionsof these units strike them as especially noteworthy. Currently we are using a cross-modal matching paradigm to replicate andelaborate on these findings. In one study, 9- to 11-month-old infants viewtwo simultaneous displays of live action, each involving a single actormanipulating a different set of objects. In one set of displays, forexample, the actor on the left places small objects in a set of drawers,while the actor on the right arranges objects on a bookshelf. While theactors are in motion, tones are presented to correspond with the completionpoints of one of the actor's intentions (e.g., a tone occurs when the actoron the left opens a drawer, places an object inside, and closes thedrawer). If infants are capable of detecting intention-relevant structurein these live action displays, they should look longer at the display inwhich the tones correspond to completion points in the actor's intentionalaction. Successful cross-modal matching in these studies requires infants tospontaneously parse a range of complex actions without the benefit ofextensive familiarization with the actions being performed. Thecross-modal matching technique thus has the potential to provide importantnew insights into the scope and power of infants' action parsing abilities.
poster
If someone is giving something to someone, the giver, the recipient, andthe transferred object are fundamental to our understanding of the causalstructure of the event. However, the clothes someone is wearing and otherirrelevant objects have a non-obligatory character in our encoding ofevents. When actions become encoded in language, the notion of what isrelevant vs. accidental is captured by the notion of argument structure. Those things that are crucial to the meaning of an event will tend to beencoded as an obligatory argument to a verb describing the action. An actof giving that does not involve a transferred object, is no longer an actof giving. Similarly, 'John gave Mary' is ungrammatical because theargument structure of the verb 'give' is incomplete.Children must learn the argument structure for each verb --whether itrequires an Agent, a Patient, a Theme, a Source, or a Goal. Deaf childrenwho invent home sign systems appear to respect argument structure even inthe absence of a language model. We propose that the acquisition ofargument structure in children requires not just hearing verbs withassociated noun phrases in the ambient language, but having the right kindof event structures encoded cognitively prior to acquiring language inearnest.To begin to test this hypothesis, we have examined whether pre-verbal 10month olds show evidence of event representations that resembled argumentstructures in language. Infants were tested in a visual habituation task. In this procedure, an infant was first habituated to a looped videosequence containing three participants. In one condition, the eventinvolved a girl giving a toy to a boy. All three participants in thisevent were obligatory. In the other condition, the event involved a girlhugging a boy while carrying a toy. In this event, only the girl and boywere required participants. The toy, while present, was not part of theargument structure of the event. After habituation, infants were testedfor the relative salience of the toy in each condition. They were exposedto a new video that contained the same action as before, but with the toymissing. As predicted, infants showed dishabituation in the GIVEcondition, but not in the HUG condition, suggesting that they only foundthe absence of the toy to be relevant when it played a fundamental role inthe action, and was ipso facto a candidate for being an verb argument in alinguistic description of the event. One possibility is that these infantsalready had acquired receptive language for the concepts of giving andhugging, and that this explains their behavior. However, we found thatinfant dishabituation in the two conditions was not associated with theirlinguistic knowledge as measured by the MacArthur CDI. This study thusprovides preliminary evidence that 10 month oldsdo encode events in amanner that could be used to bootstrap a system of argument structuresassociated with the verbs in the language.
poster
How do infants transform the ongoing activity around them into distinct andmeaningful parts? Although the individuation of actions is as fundamentalas the individuation of objects, only recently have researchers begun toexplore infants' ability to parse motion into units. Work by Wynn (1996)showed that while temporal gaps between portions of motion are very helpfulto infants in parsing motion, they are not necessary: 6-month-old infantswere able to parse an unbroken sequence of jumps interspersed with wagginginto separate, countable units. Further research (Sharon and Wynn, 1998)suggested that a repeating pattern of motion may be an important cue toinfants: in contrast to their success with the homogeneous, repeatingsequence of jumps and wags, infants failed to parse continuous,heterogeneous sequences of actions which lacked a consistent pattern (forexample, jump-fall-jump interspersed with a jiggling motion). In this study, we tested an alternative interpretation of these results:rather then relying on repeating patterns in the stimulus input, infantsmay make use of certain kinds of perceptual contrasts; specifically,tangent discontinuities (TDs) in the path of motion. A number ofresearchers (Kellman & Shipley, 1991; Shipley & Kellman, 1990) accordsimilar discontinuities a central role in object perception. The importanceof TDs arises from a basic fact of projective geometry: whenever one objectintersects with or partly occludes another, there will almost always be anabrupt change in the orientations of the projected edges; thus TDs are ahighly reliable cue to object boundaries. Applied to the individuation ofactions, this suggests that TDs in the path of motion could be an importantcue to action boundaries. For example, infants in the 'jump-wagging' studymay have individuated jumps and wagging based on the sharp change (verticalversus horizontal) in their paths of motion; but in Sharon and Wynn's(1998) experiment this cue was not reliably present, which may have led toinfants' failure in that study. To test this hypothesis, 6-month-oldinfants were presented (in the same paradigm as the previous experiments)with sequences composed of falls interspersed with a side-to-side jigglingmotion. These sequences provided a consistent, repeating pattern but lackedtangent discontinuities in the overall path of motion, as both the fallsand the jiggles began with a sideways motion. Infants failed to parse thesequences, suggesting that tangent discontinuities may in fact play animportant role in young infants' parsing of actions. This has a number ofimplications.First, it suggests a picture in which infants' ability to parse theon-going activity around them is surprisingly limited: although manyactions are signaled by abrupt changes in the direction of motion, manyother actions are not. It may be that infants initially segment only thegrossest units from an action stream and then increase the 'resolution' oftheir action parsing with increasing experience. Second, this resultraises the possibility that infants may utilize different cues for parsinganimate versus mechanical motion: for an adult, the movement of a ballwhich ricochets off a wall at a sharp angle would probably be parsed as oneaction rather than two, despite the TD in the overall path of motion. Whether infants would apply the TD principle to such a display in aquestion worthy of investigation [or, an open question]. Finally, thiswork points to the need for an appraisal of the role of TDs in infant'sparsing of objects. Although research has extensively explored thecontributions of motion and Gestalt cues to infants' object individuation,no work has explicitly evaluated the role that TDs may play. Such a studycould be highly informative.
poster
The ability to perform a particular 'means' to achieve a desired 'end' hasbeen considered by many to be the hallmark of intentional behavior ininfants (e.g. Piaget, 1953; Willatts, 1999; Carpenter et al., 1998). Assuch the ability to understand 'means-end' behavior in others forms acritical component of an understanding of intentional action. Althoughmuch is known about the development of means-end problem solving duringinfancy, little research has been directed toward when it is that infantscan interpret a means-end action sequence performed by another person asbeing goal-directed. In a recent experiment, we found that by 12-months of age, infants can usemeans-ends relations to relate distinct actions to a common goal. Infantswere presented with a means-end action sequence in which an actor placedher hand on a translucent box, opened the box and grasped a toy sittinginside the box. Having seen this event, infants later viewed the actor'sactions on the box as directed toward the toy inside the box, rather thanthe box itself, suggesting that the infants correctly recognized the goalof the action. In a follow-up condition, we disrupted the causal relationbetween the two actions so that opening the box was no longer necessary toget the toy. Infants became confused as to the goal of the actionsequence, suggesting that they appreciated the causal necessity of openingthe box to obtain the toy. That is, 12-month-olds seem to rely onmeans-ends relations to interpret action sequences.Using the same paradigm, we conducted another experiment to determinewhether younger infants could also engage in means-end reasoning aboutother people's actions. Unlike 12-month-olds, 10-month-olds could notdetermine the goal of the action sequence. Because we suspected that10-month-olds had difficulty understanding the relation between the toy andthe box, we gave them experience with these objects prior to thehabituation task. Even following this exposure, 10-month-olds did notinterpret the actor's touch of the box as a means to obtaining the toyinside the box. Currently, we are investigating whether 10-month-olds' inability to reasonabout means-end action sequences arises from a) a lack of specificknowledge about particular means-end actions (an informational deficit),or, b) a lack of means-end reasoning skills more generally (a conceptualdeficit). In this study infants see a means-end action sequence that theyare adept at performing by 8 months of age (Willatts, 1999): an actorpulling a cloth to get a toy. If 10-month-olds recognize the sequence asbeing directed toward the toy, the results may suggest that lack ofspecific information about particular means-end relations mediates theirability to make sense of particular intentional actions. In contrast, ifinfants do not recognize the goal behind the action sequence, their failureto relate means and end in the prior study may reflect a deeper, conceptualdeficit.
poster
Infants' ability to follow direction of gaze and pointing gestures hastraditionally been assessed in triadic interaction. More recently it hasbeen argued that this method of assessment may lead to an overestimation ofinfants' comprehension of referential gestures since their use of objectdirected cues my not reflect their understanding of the other person'sintentions. A method that allows to study infants' comprehension of the intentional nature of such gestures isthe visual habituation-dishabituation method. Here, infants' looking timesin reaction to expected or unexpected actions of another person aremeasured. To date few studies have used this approach. In the present studywe longitudinally assessed both infants' performance in triadicinteractions (spontaneous gaze and point following to different stationarytargets) and their comprehension of goal-directedness of pointing andlooking (compared to reaching) in a visual preference-for-novelty procedure(a modification of the procedure introduced by Woodward (1998)). Infantswere presented with a person who was seated in the middle of a stagebetween two objects. Habituation trials showed the person repeatedlygesturing (looking at, pointing at or reaching for) towards the sameobject. Then object positions were reversed and infants were shown a totalof six test trials, alternatingly the person performing the same gesturetowards the other (new) object, or gesturing to the other side towards thesame object as before. Results of the first two measurements (at 10 and 12 months of age; 15 monthassessment is under way) indicate that competence in triadic interactiondevelopmentally precedes understanding another person's referentialgestures outside an interactional context. Both, at 10 and 12 months ofage, only a minority of infants looked longer at new-object-same-path testtrials than at same-object-new-path test trials. Infants did not appear toselectively encode goal-direction in our paradigm. Rather, our findings aresimilar to those obtained by Moore (1999) with 13-month-olds when using avideo presentation. The actor's stationary posture may have led to asituation that was more similar to a video- than an interactivelive-presentation. Thus, competences may have been overestimated in moreinteractive versions of the task. Also, as opposed to the procedure used byWoodward (1998) different hands were used to point or reach to each side,so that in the new object trials the same arm and hand as duringhabituation were active, while in same object trials the other arm and handperformed the action. Thus, the failure in our task may indicate thatinfants at this age interpret individual movements of body parts asgoal-directed but do not infer a person's overarching action goal. Controlexperiments testing for these interpretations will be reported.ReferencesMoore, C. (1999). Intentional relations and triadic interactions. In P. D.Zelazo, J. W. Astington, & D. R. Olson (eds.), Developing Theories ofIntention. Social Understanding and Self Control (pp. 43-62). Mahwa, N.J.:Lawrence Erlbaum Associates. Woodward, A. L. (1998). Infants selectively encode the goal object of anactor's reach. Cognition, 69, 1-34.}