Category-Specific and Category-General Neural Codes of Recognition Memory in the Ventral Visual Pathway

The ventral visual pathway is critical for visual object perception. It has been divided into low-level and high-level visual cortices. The low-level visual cortex, which focuses more on physical properties like color, size, shape, and orientation, consists of the primary visual cortex (V1), V2 and V4 areas (Fan et al., 2021; Goddard, 2017) and is involved in the recognition of multiple categories of material, such as faces, scenes, and common objects (Taylor, 2022; Cichy et al., 2012; Sengpiel & Hübener, 1999). In contrast, brain regions in the high-level visual cortex have been suggested to be responsible for the perception of specific categories of objects (Haeger et al., 2021; Andrews et al., 2015). For example, the fusiform face area (FFA) is reported to be more sensitive to facial information than other types of information (Liu et al., 2021; Ma & Han, 2012; Scherf et al., 2011; Kanwisher et al., 1997), whereas the parahippocampal place area (PPA) is selective for images conveying place information (Silson et al., 2022; Epstein & Kanwisher, 1997).

Despite accumulating evidence for the category-general and category-specific roles of the low-level and high-level visual cortices in object perception, respectively, it is unclear whether similar roles of those brain areas play in recognition memory as in perception, given the potential different processes of perception and recognition memory (Kantner, 2019; Roussy, 2021). To address this question, previous neuroimaging studies using the “study-test” paradigm have explored the role of brain regions in the ventral visual pathway in recognition memory (Druzgal & Esposito, 2003; Köhler et al., 2002). By calculating intersubject brain-behavior correlations or comparing neural activation differences during encoding between subsequently remembered and forgotten items (i.e., subsequent memory effect, SME), researchers have revealed that both low-level and high-level visual cortices are involved in recognition memory. For example, previous studies have found that brain regions in the low-level visual cortex respond to multiple categories of materials during encoding, and neural activations in the occipital cortex during encoding predict the memory performance of scenes and common objects (Rupp et al., 2017; Sommer et al., 2005). In addition, previous studies have found SME in the occipital cortex for scenes, common objects, and faces (Hayes et al., 2017; Cansino et al., 2015; Otten, 2001).

In contrast with the involvement of brain regions in the low-level visual cortex in recognition memory of various visual materials, those in the high-level visual cortex, such as the FFA and PPA, seem to contribute to recognition memory of specific categories. Specifically, there is evidence that activations in the PPA during encoding predict memory performance of scenes (Schultz et al., 2021; Golarai et al., 2007; Brewer et al., 1998) and that activations in the FFA during encoding predict memory performance of faces (Golarai et al., 2007; Zeineh, 2003; Kuskowski & Pardo, 1999). In addition, it has been reported that the FFA and PPA showed SME for faces and scenes, respectively (Hayes et al., 2017; Liu et al., 2017; Hasinski & Sederberg., 2016; Foster et al., 2022; Prince et al., 2009; Bainbridge et al., 2017).

Taken together, the above two lines of studies have revealed that the low-level visual cortex is involved in recognition memory of various visual objects, while the high-level visual cortex (e.g., FFA and PPA) is sensitive to specific category of objects (Prince et al., 2009; Cansino et al., 2015; Hayes et al., 2017). Nevertheless, the question whether the low-level and high-level visual cortices (e.g., FFA and PPA), respectively, play category-general and category-specific roles in recognition memory of visual objects remains to be elaborated for two reasons. First, previous studies on recognition memory of faces and scenes mainly used univariate activation analysis which simply averaged the activation results of all voxels, resulting in a loss of fine-grained spatial-pattern information (Lu et al., 2020; Mur, 2009), which is thought to contain representational content (Wang et al., 2012; Mur, 2009). Therefore, multivariate pattern analysis (MVPA) that applies powerful pattern classification algorithms to extract spatial patterns from neuroimaging data is needed for investigation on the roles of regions in the ventral visual cortex in recognition memory. It should be noted that although several studies used MVPA to explore the activation patterns of recognition memory, they mainly focused on regions in medial temporal lobes (i.e., hippocampus, perirhinal cortex, and parahippocampal cortex) rather than those in the ventral visual pathway (Martin et al., 2013; Martin et al., 2016). More importantly, in contrast with univariate analysis, MVPA has an advantage in addressing the question about category-general and category-specific neural codes in recognition memory because this method can be used to make both within-category and cross-category predictions based on the fine-grained activations of multiple voxels (Haxby, 2001; Haynes, 2015). If both within-category and cross-category predictions are significant, the brain region carries category-general neural codes of recognition memory for different materials. If only the within-category prediction is significant, the brain region plays a category-specific role in recognition memory.

To address the above questions, the present study adopted a subsequent memory paradigm and MVPA to explore category-specific and category-general neural codes of recognition memory in the ventral visual pathway. Twenty-five participants performed a localizer task and two encoding tasks (i.e., face and scene encoding) during scanning and then performed a recognition test 30 minutes after the fMRI scan. We first defined the right FFA and bilateral PPA in the high-level visual cortex and the bilateral inferior and superior occipital cortex in the low-level visual cortex as regions of interest (ROIs). Then, we examined the SME in those ROIs by contrasting neural activations of remembered and forgotten items. Finally, we performed both within-category MVPA and cross-category MVPA in those ROIs. Based on previous findings that the low-level and high-level visual cortices are differentially involved in recognition memory of visual objects, we hypothesized that the former brain regions would carry category-general neural codes of recognition memory, but the latter ones would play a category-specific role in recognition memory.

留言 (0)

沒有登入
gif