The neural infrastructure supporting visual processing in humans has been extensively investigated resulting in a hierarchically-organized model of object processing in ventral occipitotemporal cortex that also supports the semantic processing of visual objects. Such a model has been combined with cognitive models of semantic knowledge leading to accounts of how semantic information is instantiated in the brain. However, little is known about how meaningful information evolves during time, from visual inputs to a fine-grained understanding of what the object is. Addressing this central issue motivates the experiments reported in this thesis. Specifically, I ask how meaning emerges from perceptual inputs, what kind of semantic measures meaning is expressed through, and what neural mechanisms integrate meaningful information to form coherent conceptual representations. I used behavioural and neuroimaging techniques with advanced analysis methods in a number of object naming experiments. I used time-sensitive magnetoencephalography (MEG) to record the temporal dynamics of neural activity during object naming, allowing (a) perceptual and semantic measures to be correlated with the MEG data and (b) isolating different aspects of perceptual and semantic processing. I also used MEG to ask what neural mechanisms afford the integration of semantic information by manipulating the types of semantic information required for subjects to name pictures. In a third experiment, I used visual masking to manipulate the amount of conceptual processing that could occur, to test if certain aspects of conceptual representations are available prior to others. Results provide converging evidence that conceptual representations are rapidly accessed from perceptual inputs and evolve along a coarse-to-fine trajectory. The extraction of meaning from visual objects is shown to rely on dynamic, recurrent processes that function to integrate the more subtle aspects of the emerging conceptual representation.