From birth, infants have to coordinate vision and action to explore their environment. Around 10 months of age, they start reaching for out-of-reach objects by pulling a string attached to them, or a cloth on which the objects stand. This type of behavior, called means-end behavior, involves a key concept: the notion of spatial connectedness. Psychologists have shown that the presence/absence of a spatial gap between objects influences the performance of infants in such means-end behaviours. Infants are able to identify composite objects as a unique object when both are contiguous and move in a similar way. However, little is known about what infants perceive and understand from the spatial relationship between unmoving objects. When do infants start to consider the spatial connection as a relevant information to identify composite objects ? And then, as soon as they understand that two contiguous objects are connected to each other, how do they apply these informations to solve problems involving the retrieval of out-of-reach objects ? Are these informations sufficient in situations with more complex spatial relationships ? Two behavioral studies [one involving an eyetracker] on infants aged 6 to 20 months gives us more informations on infants' expectations of composite objects, and their perception of spatial connectedness.