3.1. Formation and generalization of visual representations of new 3-D objects
A major function of the human visual system is to afford reliable recognition of three-dimensional (3D) objects from two-three-dimensional retinotopic inputs. Influential theories have been proposed to explain how visual representations may allow rapid and accurate visual recognition. However, processes underlying the formation of these representations remain unclear. Intriguingly, although the hippocampus is critically involved in memory processes, its role in visual memory and visual learning is not well understood. Here, we first aimed to investigate the neural bases of 3D-object learning by assessing changes in brain activity during successive periods of training on unfamiliar 3D-objects. A second main goal was to test whether learning-related activation strength in any of the contributing regions predicted how well subjects could generalize 3D-object knowledge across different views. Finally, because accurate recognition of our 3D-objects involves the use of spatial cues (i.e., tridimensional arrangement of the objects’ parts), we investigated whether 3D-object learning might be related to more general spatial abilities.
More specifically, this study tested the following hypotheses:
H1.1. View-specific and view-invariant patterns of activity coexist in the brain and allow 3D objects recognition
H1.2. Regional brain activity during learning might predict performance at generalizing across new views of the previously learned objects
H1.3. 3D-object learning correlates with spatial abilities
We expected a distributed network of brain areas, including among others, the lateral occipital cortex, the fusiform gyrus, the hippocampus, to be involved in the learning and recognition of 3D-objects. We also expected to find both view-specific and view invariant patterns of activation across the network. We predicted that the hippocampus may play a crucial role in the structural learning of these objects, and that its activity at learning may predict performance at generalizing across novel views. Finally, we expected performance at this task to be correlated with more general abilities, such as mental rotation and/or navigation.
3.2. Role of attention at encoding on subsequent explicit and implicit memory
The brain extracts relevant information from the environment depending on immediate goals and expectations. The second part of the present thesis is based on the postulate that this selective process might impose major constraints on subsequent memory traces. As described in section 2.4.4, attention at encoding may be necessary for the formation of memory traces that can be accessed explicitly. By contrast, some perceptual mnesic traces might be created for ignored stimuli and be accessed implicitly. However, while the effect of selective attention has been clearly shown for explicit memory, doubts remain on the exact role of attention on implicit memory. The main goal of the second study presented in this thesis was to investigate the effect of selective attention at encoding on subsequent implicit memory, and thus to assess the nature of the representations created for attended and ignored stimuli (abstract vs.
perceptual).
The study was conducted to test the following hypotheses:
H2.1: attention at encoding is mandatory to create durable explicit memory H2.2: durable implicit memory traces can occur without attention at encoding H2.3: implicit and explicit memory tests involve different levels of representations
Predictions were that explicit memory, assessed by an old-new recognition task, would be critically affected by the lack of attention at encoding, whereas some implicit memory traces, measured through a fragmented-picture completion task, may arise for ignored stimuli. We also expected attention at encoding to lead to the formation of abstract representations in explicit memory, and more perceptual ones in implicit memory.
The results are described in Section 3 of the Experimental Part.
This study has been extended at the neural level in a complementary fMRI study by Vuilleumier et al. (Vuilleumier et al., 2005). The results are summarized at the end of concerned section and details can be found in the annexed article (annex 1).
3.3. Role of sleep in memory for face: behavioral investigations
Faces are very homogeneous stimuli, associated to high social value, and thus, are of particular relevance while studying the formation of individual memory traces. Although the main theories of person identity recognition offer some predictions about face priming and recognition (see section 1.2.1), how new face representations are built, altered, or refined over time to allow for efficient recognition of face identities is still poorly understood. Based on increasing evidence for a role of sleep in learning and memory consolidation (see section 2.5.2), we hypothesized that nocturnal sleep might provide a permissive condition for the occurrence of consolidation of new face representations.
More specifically, the hypotheses tested in this study were the following:
H3.1: sleep has an effect on both implicit and explicit memory of faces
H3.2: a sufficient level of encoding is required to allow memory consolidation to occur
H3.3: sleep consolidates mnesic traces by making them more abstract and resistant to visual input conditions
We expected that implicit memory measured by reaction times (RT) repetition priming and mere-exposure effects would be more important when study and memory-test were separated by a delay including a period of sleep rather than a day of wakefulness. We also predicted that similar effects would occur for explicit memory, measured by an old-new recognition task.
We expected the level of encoding to be critical in this condition. Finally, we expected that memory for faces after a period of sleep would be more resistant to perceptual changes (such as size).
The results are described in Section 4 of the Experimental Part.
3.4. Role of sleep in memory for faces: an fMRI investigation
Investigating the role of sleep on explicit memory using fMRI is particularly interesting since it allows to test for the role of hippocampal circuitry, known to be involved in sleep-related memory consolidation (e.g., Peigneux et al., 2004; Rasch et al., 2007). However, this effect remains unclear in face processing, and we don’t know how sleep might modulate face representations. Moreover, recent work on word recognition shows that sleep might prevent false memory. Face recognition rely on a distributed neural network, including the Fusiform Face Area (FFA), as well as the medial temporal lobe and frontal areas (see section 1.2.2).
The fusiform face area might be the storage site of face representation and might process configural information. The medial temporal lobe, generally involved in memory, may be the key structure involved in the comparison of the input and the stored representation and, finally, frontal areas may be involved in retrieval strategies, and protection against false recognition. To date, there is no fMRI study on the role of sleep on face memory. The present study investigates the role of sleep on face memory consolidation, and the probable modulation of neural activity in face-specific and memory-related areas.
We hypothesized that:
H5.1: sleep following exposure to new faces may improve explicit memory for faces
H5.2: sleep-related explicit memory enhancement relies on the strengthening of configural information
H5.3: the effect of sleep on face memory is associated with the modulation of activity in the hippocampus and in face-responsive brain areas.
We predicted that sleep would enhance recognition for previously learned faces, and would increase the robustness of memory (i.e., better resistance towards distracters), probably by strengthening configural memory. We also expected the hippocampus and fusiform face area to be related to this enhancement of memory.
The results are described in Section 5 of the Experimental Part.
EXPERIMENTAL PART