Inferring the Size of a Goal Object from an Actor's Grasping Movement in 6- and 9-Month-Old Infants

The present study applied a preferential looking paradigm to test whether 6‐ and 9‐month old infants are able to infer the size of a goal object from an actor's grasping movement. The target object was a cup with the handle rotated either towards or away from the actor. In two experiments, infants saw the video of an actor's grasping movement towards an occluded target object. The aperture size of the actor's hand was varied as between‐subjects factor. Subsequently, two final states of the grasping movement were presented simultaneously with the occluder being removed. In Experiment 1, the expected final state showed the actor's hand holding a cup in a way that would be expected after the performed grasping movement. In the unexpected final state, the actor's hand held the cup at the side which would be unexpected after the performed grasping movement. Results show that 6‐ as well as 9‐month‐olds looked longer at the unexpected than at the expected final state. Experiment 2 excluded an alternative explanation of these findings, namely that the discrimination of the final states was due to geometrical familiarity or novelty of the final states. These findings provide evidence that infants are able to infer the size of a goal object from the aperture size of the actor's hand during the grasp.

The development of the ability to perceive and understand actions as goal‐directed has become more and more a central issue in the field of social‐cognitive development in recent years. Major advances have been achieved not only in infancy research ([1]; [6]; [13]; [30]) but also in research on human cognition per se (e.g. [15]).

Recent research suggests that the ability to understand goal‐directed actions and the intentions of an acting person might be a precursor of a 'theory of mind' in infancy ([1]; [27]; [28]). Starting at the age of 6 months, infants become able to perceive and interpret a human action as goal‐directed ([19]; [29]). To perceive an action as goal‐directed, it can be performed either by a human agent ([14]; [18]; [29]) or by a non‐human agent ([20]). In these studies, the presented action was always geared towards a visible object, it was completed and infants could perceive the achievement of the goal of the action. Such an event shown was, for example, an actor's hand grasping one of two objects ([29]), pushing the object to a different location ([18]), or an inanimate object moving to and arriving at one of two objects ([20]). Fewer studies have investigated the perception of uncompleted goal‐directed actions. In imitation studies, 15‐ and 18‐month‐olds were shown to infer an adult's intended act by watching failed attempts of the actor, in which an intended goal was not fulfilled ([17]; [21]). In looking time studies using inanimate objects, inference of goal‐directedness is already evident in 12‐month‐olds ([7]). Based on these findings, it might be interesting to ask if infants in their first year of life are already able to infer the goal of an uncompleted reaching action performed by a human agent.

In two recent studies, the perception of goal‐directed but uncompleted reaching actions was investigated in young infants using a preferential looking paradigm ([8]) and an imitation paradigm ([12]). In both studies, infants first saw the uncompleted reaching action towards one of two objects and were then either presented simultaneously with two final states of the action ([8]) or reached towards the same objects the actor was reaching to previously ([12]). Both studies showed that already at the age of 6 to 7 months, infants were able to encode the goal of an action, which was not actually perceivable. However, neither study ruled out the possibility that infants' behavior was just based on a simple path extrapolation strategy instead of goal inference. A further critical point that applies not only to these two studies ([8]; [12]) but to all other studies investigating the perception and encoding of goal‐directed behavior in young infants is that the goal objects were always visible during the presentation of the action and the test events. The differentiation between two final states or the selection of an action might therefore be based on an actor–object relation without any inference about goal‐directedness (e.g. [5]).

The present study was designed to further investigate infants' perception and interpretation of goal‐directed but uncompleted manual actions. The design used in previous studies was improved in the following respects: First, we were interested in whether infants are able to encode the actor's selection of one of two target actions towards the same object, and not whether infants are able to encode the actor's selection of one of two target objects. As a consequence, only one target object was used instead of two. And second, to exclude the possibility that infants' looking behavior is based on actor–object relations instead of inference about goal‐directedness, the target object was not visible during the first part of the grasping action. For this purpose, a perception task was conducted in which an object‐related grasping movement1 was presented to the infants with an actor either grasping with a large hand aperture size (as for a large object) or grasping with a small hand aperture size (as for a small object) with the target object not being visible during the grasping movement (see Figure 1a). In a preferential looking paradigm, we then tested whether infants at the age of 6 and 9 months inferred that the goal of an actor grasping with a large hand aperture size is to grasp a large object whereas the goal of an actor performing a grasp with a small hand aperture size is to grasp a small object. Thus, we were interested in whether infants are able to infer the size of the target object based on the size of the hand opening during grasping.

Graph: 1 (a) Grasping movement with large (left panel) and small (right panel) hand aperture size, (b) Final states of Experiment 1, (c) Final states of Experiment 2.

The purpose of Experiment 1 was to test whether 6‐ and 9‐month‐old infants are able to infer the size of a goal object from an actor's uncompleted grasping movement. The stimulus presentation consisted of a video display of an actor grasping towards a cup with the handle showing either towards the actor or away from him. The actor performed a grasp with either a large hand aperture size or a small hand aperture size (see Figure 1a). The cup was occluded during the first part of the grasping movement and the actual achievement of the goal (the grasp) was not presented to the infants. Subsequently, in a preferential looking paradigm, the static images of two final states of the grasping movement were presented simultaneously on two separate monitors (forced choice preferential looking paradigm, FPL). The expected final state showed the actor's hand holding the cup in a way that would be expected based on the size of the hand aperture during the observed grasping movement (for details see Figure 1b and description below). In the unexpected final state, the actor's hand held the cup at the side, which would be unexpected based on the size of the hand aperture during the observed grasping movement. Parallel presentation of the two final states was preferred to successive presentation because when using parallel presentation of two stimuli, infants have to rely less on stored memories, but can directly compare the two outcomes of an event. Moreover, the design of the study is simplified and order does not have to be counterbalanced (see also [8]).

Thirty‐two 6‐month‐old infants (19 girls, 13 boys; mean age: 6 months; 5 days, range: 5;19–6;14) and 32 9‐month‐olds (12 girls, 20 boys; mean age: 9;3, range: 8;17–9;15) participated in Experiment 1. Seventeen additional 6‐month‐olds (ten girls, seven boys) were tested but not included in the final sample due to distress or fussiness (n = 14), or technical problems (n = 3). Twenty‐three additional 9‐month‐olds (16 girls, seven boys) were tested but not included in the final sample due to distress or fussiness. Infants' names were obtained from public birth records. The large number of infants tested was intentionally chosen in order to compensate for a potential bias that the position of the test stimuli might be confounded with an individual infant's baseline preference. Due to this number of tested infants, possible individual baseline preferences are very likely to cancel each other out.

The laboratory was an unfurnished room except for the test equipment. Infants were seated in a safety car seat (Maxi Cosi Cabrio), which was brought to an upright position by a wooden sub‐construction.

The stimuli were presented on three monitors (Neovo LCD Display X19AV) via three DVD players (Cyberhome CH‐DVD 462). The monitors were arranged in a pyramidal way, one monitor on the top, two monitors at the bottom. The viewing distance between this experimental setup and the children was 80 cm. At the beginning of the experiment, on all three monitors the picture of a blue curtain was presented for 10 s. Then, the attention of the infant was drawn to each of the three monitors by presenting a red animated and smiling face with a sounding tone on each monitor successively, with the picture of the blue curtain still being presented on the two other monitors (order of sequence: upper monitor – lower left monitor – lower right monitor). This sequence also served as a calibration in the offline scoring of infants' looking behavior. During stimulus presentation every trial started with a fresh presentation of the attention grabber on the upper monitor. Then, still on the upper monitor, the two hands of a male adult (actor) were presented both lying flat on a table. In front of the actor's hands, a grey occluder (21 cm × 29.5 cm) was visible, 17 cm distance from the actor's hands. After a 1.3 s still phase, the actor performed a one‐handed grasping movement with the right hand. The hand opening during the grasping movement varied as a between‐subjects condition. In the large aperture size condition, the hand opening was wide as if the actor was grasping for a large object (see Figure 1a, left panel). In the small aperture size condition, the hand opening was small as if the actor was grasping for a small object (see Figure 1a, right panel). This grasping movement was presented until the actor's fingertips just reached the occluder (see Figure 1a, both panels), then the video on the upper monitor stopped. The duration of this first part of the grasping movement amounted to 0.7 s. As this was a rather short time for the infant to acquire exact information from the scene, the movement was followed by a fixed‐image of the final position of the grasping movement for 0.8 s. Then, the blue curtain was presented on the upper monitor until the end of the trial. Subsequent to the end of stimulus presentation on the upper monitor, the freeze frames of two final states of the grasping movement were presented simultaneously for 20 s on the two lower monitors. Here, the occluder had been removed and the actor's hand holding a cup with the handle pointing either towards the actor or away from him was presented. In the expected final state, the actor's hand was holding the cup in a way that would be expected based on the size of the hand aperture during the observed grasping movement (i.e. holding the cup at the handle after performing a grasp with a small hand aperture size or holding the cup on the other side after performing a grasp with a large hand aperture size, see Figure 1b). In the unexpected final state, the actor's hand held the cup at the side, which would be unexpected based on the size of the hand aperture during the observed grasping movement (i.e. holding the cup at the handle after performing a grasp with a large hand aperture size or holding the cup on the other side after performing a grasp with a small hand aperture size; see Figure 1b). A total of six identical trials following the described pattern were presented to the infants. The following stimulus variations were counterbalanced between subjects: Grasp size (large aperture size vs. small aperture size), target color (green cup vs. red cup), and position of expected and unexpected final state (expected final state on the lower left monitor and unexpected final state on the lower right monitor and vice versa). The infant's looking behavior was recorded using a small camera (board camera, VK 1312), which was positioned between the three monitors.

Infants were tested in the laboratory at a time of day when they were likely to be alert and in a good mood. All infants were tested individually with one parent present. Each participant and parent were first escorted to a reception room. For approximately 10 minutes the infant was allowed to explore the room, while the research assistant described the test procedure to the parent. The infant and their parent were then brought to the test room. The research assistant helped the parent to position the infant in the safety car seat. During stimulus presentation, the parent sat on a chair behind the safety car seat. Parents were instructed not to interact with their children during testing. They were encouraged, however, to put both hands symmetrically close to the child if it appeared necessary to comfort the infant. Once the infant and the parent seemed comfortable, the research assistant left the room and the stimulus presen‐tation was started.

Looking times were coded from video by a trained observer using the software INTERACT (Mangold Software & Consulting GmbH, Arnstorf, Germany). A total of three trials were analyzed per infant. The very first trial of the sequence was not included in the analysis to provide the infant with an introductory trial in which an infant is able to get used to the two final states of the action and to orient him‐ or herself to what is actually presented in these test events. Of the remaining five trials the first three trials were included in the data analysis in which (a) the infant had attended to the first part of the action sequence presented on the upper monitor, (b) the infant did not show any signs of fussiness during the presentation period of the two final states, and (c) the parents did not interfere with the stimulus presentation. The number of three trials was chosen for the following reason. The presentation period of the two final states was 20 s in which no action was presented to the infants. This rather boring test phase caused quite a bit of fussiness leading to a number of trials that could not be included in the final sample. To avoid testing a huge amount of infants in order to get all trials included in the final sample, we decided to integrate three valid trials, which was the number that the majority of infants achieved. The decision whether a trial was included in the final sample or not was made prior to the examination of the participant's data. One trained observer scored all trials, and as a check on the reliability of scoring, a second observer scored a random sample of 25% of the participants. The correlation of agreement was r =.97 for both the 6‐month‐olds and the 9‐month‐olds.

For the main analyses, the total amount of looking time on the three valid trials was calculated for all infants. A preliminary analysis of variance with the between‐subjects factors sex, target color, hand aperture size during the grasping movement, and position of expected and unexpected final state yielded no significant main effect on infants' total looking time (all Fs < 1). For subsequent analyses, data were collapsed across groups.

The main findings of Experiment 1 are displayed in Figure 2 (left panel). A 3 × 2 × 2 (Trial × Final State × Age) repeated measures ANOVA was performed on infants' mean looking times with trial and final state (expected vs. unexpected) as within‐subject factors and age as between‐subjects factor. Infants looked longer at the unexpected final state than at the expected final state, F(1, 62) = 10.11, p <.01 (unexpected: M = 7.38 s, SD= 3.14 s; expected: M = 5.52 s, SD = 2.44 s). This main effect was independent of age, F < 1. There was a main effect of age, F(1, 62) = 4.03, p <.05, with the 9‐month‐olds looking longer at the two stimulus displays (expected + unexpected final state; M = 13.67 s, SD = 2.73 s) than the 6‐month‐olds (M = 12.13 s, SD = 3.37 s). No further main effect or interaction was significant (all ps >.19). The main effect of final state was confirmed by a nonparametric analysis: 41 infants looked longer at the unexpected final state, and 23 infants looked longer at the expected final state (Sign test, p <.05).

Graph: 2 Total amount of infants' looking times over three consecutive trials from Experiment 1 and Experiment 2 (with standard error bars).

One might raise the concern that the initial sequence of the attention grabber might have biased the infants to look according to this ordering, thus, looking on the left monitor first, then looking on the right monitor. In order to test this, we additionally analyzed both where the infants first looked during the first trial and during the trials that were included in the final analysis. During the first test trial, 35 of the infants first looked towards the lower right monitor, 29 of the infants towards the lower left monitor (one‐way chi‐square: χ2(1, _I_N_i_ = 64) = 0.56, p =.53). The looking times towards the left and right monitors were compared using a one‐way ANOVA. Infants looked equally long at the two displays, F(1, 63) = 2.03, p =.16 (left: M = 6.01 s, SD = 2.76 s; right: M = 6.89 s, SD = 3.10 s).

One might further raise the concern that the features of the final states might mediate looking times independently of object inferences, and a baseline preference for one stimulus could inflate the main effect of final‐state consistency despite counterbalancing. In order to test this, we additionally analyzed the looking times to either final state independent of their expectancy status by using a one‐way ANOVA. Infants' looking times did not differ significantly between the hand with the small aperture size and the hand with the large aperture size, F(1, 63) = 2.48, p =.12 (small aperture: M = 6.94 s, SD = 2.84 s; large aperture: M = 5.96 s, SD = 3.01 s).

In Experiment 1, infants aged 6 and 9 months looked significantly longer at the final state of an object‐related human grasping action which was unexpected with respect to the aperture size of the hand during the previously presented grasping action. Thus, already at the age of 6 months, infants seem to be able to infer the size of a goal object from an actor's grasping movement itself without actually seeing the target object. However, before we draw this conclusion we first have to rule out an alternative explanation of the results. Infants' discrimination between the expected and the unexpected final state could be the result of a simple comparison of the geometrical relations of the final position of the hand aperture size during the grasping movement and the opening of the hand in the final state. After a grasping movement with a large hand aperture size, for example, the expected final state would be the actor holding the cup at the large side. However, the hand holding the cup at the large side is geometrically more similar to the previously presented end of the grasping movement than the hand holding the cup at the handle. The longer looking times could, therefore, be caused by this geometrical congruency and not be due to an understanding of goal‐directed actions. This alternative explanation was tested in Experiment 2.

In Experiment 2, 6‐ and 9‐month‐old infants were presented with the same object‐related grasping movement displayed in Experiment 1 but with different final states. In these final states, the hand was placed in the same position as in Experiment 1 but without the cup being present (see Figure 1c). If geometrical congruency caused the results obtained in Experiment 1, then infants should exhibit the same looking behavior independent of whether the cup is present or not.

Thirty‐two 6‐month‐old infants (12 girls, 20 boys; mean age: 6 months; 4 days, range: 5;19–6;14) and 32 9‐month‐olds (17 girls, 15 boys; mean age: 9;1, range: 8;21–9;14) participated in the experiment. Ten additional 6‐month‐olds (seven girls, three boys) were tested but not included in the final sample due to distress or fussiness (n = 8) or interference by the parent (n = 2). Seventeen additional 9‐month‐olds (eight girls, nine boys) were tested but not included in the final sample due to distress or fussiness (n = 14), interference by the parent (n = 1) or technical problems (n = 2). Infants' names were obtained from public birth records.

The same apparatus was used to generate the stimulus display as in Experiment 1 except for the following modifications. No target object (i.e. the cup) was present during the presentation of the two final states. Care was taken that the position of the hand was identical to the stimulus presentation of Experiment 1 where the actor held a real cup in his hand (see Figure 1). The procedure and the analysis of looking times were identical to Experiment 1. The correlation of agreement between the two observers was r =.97 for both the 6‐month‐olds and the 9‐month‐olds.

For the main analyses, the total amount of looking time on the three valid trials was calculated for all infants. As a preliminary analysis of variance yielded no significant effects of the factors sex, hand aperture size during the grasping movement, and position of the expected and the unexpected final state (all p‐values >.19), data were collapsed across these factors for subsequent analyses.

The main findings obtained in Experiment 2 are displayed in Figure 2 (right panel). A 3 × 2 × 2 (Trial × Final State × Age) repeated measures ANOVA was performed on infants' mean looking times with trial and final state (expected vs. unexpected2) as within‐subject factors, and age as a between‐subjects factor. This analysis yielded no significant main effect of expectancy of display, F(1, 62) = 1.89, p =.17. There was no main effect of age, F(1, 62) = 2.41, p =.13, but age interacted marginally with expectancy of display, F(1, 62) = 3.05, p =.09. Paired‐sample t‐tests indicated no difference in looking times for the 6‐month‐olds (expected final state: M = 5.99 s, SD = 3.16 s, unexpected final state: M = 6.18 s, SD = 2.40 s; t(31) = 0.25, p =.80), but the 9‐month‐olds looked reliably longer at the expected final state (expected final state: M = 6.26 s, SD = 2.64 s, unexpected final state: M = 4.62 s, SD = 2.39 s; t(31) = 2.27, p <.05). No further main effect or interaction was significant (all ps >.13). The findings of the expectancy of display were supported by a nonparametric analysis; overall, 39 infants looked longer at the expected final state (14 6‐month‐olds, 25 9‐month‐olds) and 25 infants looked longer at the unexpected final state (18 6‐month‐olds, seven 9‐month‐olds) yielding a non‐significant effect in the overall Sign test, p =.10, but a significant effect in the 9‐month‐olds (p <.01; 6‐month‐olds: p =.60).

This rather counterintuitive finding obtained in the group of 9‐month‐olds was further investigated using a 2 × 2 (Final State x Grasp Size) repeated measures ANOVA with the aperture size of the hand during the prior grasping movement as a between‐subjects factor. This analysis yielded no significant interaction with final state, F < 1.

Similar to Experiment 1, in order to test whether the initial sequence of the attention grabber might have biased infants' looking behavior, we additionally analyzed both where the infants first looked during the first trial and during the trials that were included in the final analysis. During the first test trial, 36 of the infants first looked towards the lower right monitor, 28 of the infants towards the lower left monitor (one‐way chi‐square: χ2(1, _I_N_i_ = 64) = 1.00, p =.38). The looking times towards the left and right monitors were compared using a one‐way ANOVA. Infants looked equally long at the two displays, F(1, 63) < 1 (left: M = 6.01 s, SD = 2.71 s; right: M = 5.52 s, SD = 2.72 s).

And again, we analyzed the looking times to either final state independent of their expectancy status by using a one‐way ANOVA. Infants' looking times did not differ significantly between the hand with the small aperture size and the hand with the large aperture size, F(1, 63) = 3.70, p =.06 (small: M = 5.26 s, SD = 2.58 s; large: M = 6.27 s, SD = 2.78 s). Please note that if we look at the (non‐significant) differences between the final states, discrepancies between the two experiments can be found (a tendency to look longer towards the hand with small aperture size in Experiment 1 and a tendency to look longer towards the hand with large aperture size in Experiment 2).

Experiments 1 and 2 were additionally compared using a 3 × 2 × 2 × 2 (Trial × Final State × Age × Experiment) repeated measures ANOVA. This analysis yielded a significant effect of Experiment, F(1, 124) = 5.90, p <.05. Infants' overall looking times were longer in Exper‐iment 1 than in Experiment 2. Experiment interacted significantly with age, F(1, 124) = 6.26, p <.05, and this difference was larger in 9‐month‐old than in 6‐month‐old infants. A further analysis showed that the overall looking time was equal for both experiments in 6‐month‐olds (t < 1) and was less in Experiment 2 than in Experiment 1 in 9‐month‐olds, t(62) = 3.89, p <.001. Experiment further interacted with expectancy of display, F(1, 124) = 10.83, p <.01, the difference between the looking time toward the expected final state and the unexpected final state was significant in Experiment 1 but not in Experiment 2. No further interaction with Experiment was significant (all ps >.09).

The present results show that the findings obtained in Experiment 1 cannot be explained by a simple geometrical congruency strategy. When the grasping action was not goal‐directed towards an object but only ended with different positions of the grasping hand, the 6‐month‐olds did not discriminate between the two final states and the 9‐month‐olds even looked longer at the expected final state than at the unexpected, which is in contrast to the original findings. The diverging results found in Experiments 1 and 2 in the 6‐month‐olds cannot be explained by differences in looking times and, therefore, be a result of a different encoding of the scene as they looked equally long in both experiments.

The reversed finding in the looking times of the 9‐month‐olds compared to Experiment 1 is somewhat more difficult to explain. It does not reflect the preference for a grasping movement ending with a closed hand if no target object is present (which would be a somewhat 'natural' endstate). Such a preference would have caused longer looking times towards the large grasp final state in both conditions causing an interaction of the factors Grasp Size and Expectancy of Display, which was not found in the statistical analysis of the results. The preference for the 'expected' final state might reflect a preference for similarity. As the action has no graspable object as a goal, and thus no expected or unexpected final state to discriminate, infants might prefer the display which somehow resembles the hand aperture size during the grasping movement.

In two experiments, infants' ability to infer the size of a goal object from an actor's object‐related but uncompleted grasping movement was investigated using a preferential looking paradigm. Six‐ and 9‐month‐old infants first watched the video of an actor grasping towards an occluded cup with either a large hand aperture size or a small hand aperture size and subsequently saw an expected and an unexpected final state of this grasping action. Results from Experiment 1 showed that both 6‐ and 9‐month‐olds discriminated between the two final states and looked reliably longer at the unexpected final state than at the expected final state. Experiment 2 controlled for an alternative explanation of the findings obtained in Experiment 1, that the difference in looking times could be due to a simple strategy of geometrical familiarity or novelty of the grasping movement and the final state. In this experiment, no target object was pre‐sented. The 6‐month‐olds did not differentiate between the two final states and the 9‐month‐olds even looked longer at the expected final state.

One might argue that it is unclear how the presence and the absence of the cup in the two experimental conditions may have altered the perception of the final states between the two experiments. The cup being present in Experiment 1 may have provided some sort of reference for analyzing the hand in Experiment 1 compared to Experiment 2 where the attention of the infants may have been altered due to the absence of the cup. Indeed, removing the cup in Experiment 2 might have yielded a different gist of the goal of the perceived action. In our opinion, the different pattern of results between the two experiments is a consequence of the different object affordances. We try to illustrate this in the following: (a) Hands are very salient and important 'objects' in the infant's world. Infants show an early understanding of actions being performed by a human hand by discriminating between an unexpected and an expected test event ([29]). However, if the same action is performed by a hand that is covered with a glove or by a mechanical claw then infants do not discriminate between the test events ([11]; [18]; [29]). Thus the presentation of an action performed by a human hand is very salient and is probably processed differently than an action performed by another agent. In our study we used the presentation of a hand in both experiments, thus the saliency and the underlying processes are likely to be very similar. (b) The presentation of the grasping action in the first part of each trial was identical in both experiments. Infants consequently built the same expectation about the continuation of the action in both experiments. (c) The main purpose of Experiment 2 was to test the alternative explanation of the findings of Experiment 1, that is, that the findings may be based on a geometrical congruency effect and not on an encoding of the goal of an action. If this was the case, then the same pattern of results would have occurred in Experiment 2 as in Experiment 1. The results of Experiment 2, however, show that this is not the case and therefore rule out the alternative explanation of the findings of Experiment 1 that the differences in looking times are based on a geometrical congruency effect. The presence of a goal or target object seems to cause a differential processing of the perceived action beyond comparing geometrical congruencies.

The results of the present study therefore indicate that infants, beginning at least at the age of 6 months, are able to infer the size of a goal object from an actor's grasping movement even when the target object which is grasped for is not visible during the action. The only information available during the actor's grasping movement was the aperture size of the hand opening. Thus, to encode the actor's goal, infants have to infer the size of the target object from the aperture size of his hand during the grasping movement. These findings replicate and extend findings obtained in previous studies indicating (1) that 6‐month‐old infants are able to encode the goal of an uncompleted reaching action geared towards one of two target objects ([8]) and (2) that 7‐month‐old infants understand unfulfilled grasps and direct their own action at the goal ([12]). Unlike in these studies, the target object was not visible during the actor's goal‐directed movement in the present study and, moreover, infants had to encode the actor's selection of one of two possible target actions towards the same object. Thus, the results indicate that infants are able to infer the size of the target object from the aperture of the actor's hand during the first part of the grasping movement. The results cannot be based on a simple path extrapolation strategy as this pattern of results is dependent on the existence of a target object. It is more likely that infants used the aperture size of the hand to anticipatorily infer the goal of the action. This goal anticipation creates an expectation about the size of the grasped target and this expectancy is fulfilled in one final state and is violated in the other final state.

One further and more speculative conclusion of the present findings is that infants do not necessarily seem to need to be able to perform an action in order to understand the same action and to infer the goal of this action. Infants start to adjust their hand aperture related to the size of the target object beginning at the age of 9 months ([26]). Younger infants at the age of 6 months do not yet show this anticipatory control of their grasping movement.3 In the present study, however, infants already discriminated at the age of 6 months between the differently expected final states of a grasping action. This result is in line with several findings from studies showing that infants' performance in a visual discrimination or perception task is not necessarily based on their competence and performance in a related action task. In studies on the A‐not‐B‐error task, 8‐month‐old infants succeed in a visual discrimination task ([2]), while infants of the same age fail in an active searching task ([10], but see also [3]). Similarly, studies on solidity and continuity showed that infants already succeed in a visual discrimination task at the age of 3 months ([25]), while toddlers at the age of 2.5 years failed in a similar search task ([4]). And, finally, in a study using a perception and an action version of a means–end task, 6‐month‐olds were able to understand goal‐directed behavior in a perception task, which was independent of their performance in the action task ([9]; but see [24]).

This conclusion has, however, to be drawn with great caution. The comparison of the findings of infants' performance on an object‐related and goal‐directed grasping action ([26]) and the present findings on the perception and interpretation of a similar action performed by another person is still based on a between‐subjects comparison. For the present study, it was just assumed that only the 9‐month‐olds are able to adjust their hand in anticipation of an object, whereas the group of 6‐month‐olds lacks this experience. However, research on the development of prehension in infancy has yielded inconsistent results on the exact age at which different grasping skills emerge (e.g. [22]; [23]).

For future research it would therefore be very interesting to directly compare infants' grasping performance with their perception and understanding of grasping movements performed by another person in a within‐subject paradigm.

Parts of these findings were reported at the Biennial Meeting of the Society for Research in Child Development in Boston, USA, March/April 2007. The authors would like to thank Paul C. Quinn as well as two anonymous reviewers for constructive and helpful comments on an earlier version of the manuscript. We would like to thank the parents and infants who participated in this study. We also wish to thank Jana Hiller for technical support, Gabi Karn for the acquisition of the infants, and Petra Schradi as well as our student research assistants for help in data collection and scoring the video tapes.