Limits...
Supranormal orientation selectivity of visual neurons in orientation-restricted animals.

Sasaki KS, Kimura R, Ninomiya T, Tabuchi Y, Tanaka H, Fukui M, Asada YC, Arai T, Inagaki M, Nakazono T, Baba M, Kato D, Nishimoto S, Sanada TM, Tani T, Imamura K, Tanaka S, Ohzawa I - Sci Rep (2015)

Bottom Line: Our results demonstrate that restricted sensory experiences can sculpt the supranormal functions of single neurons tailored for a particular environment.The above findings, in addition to the minimal population response to orientations close to the experienced one, agree with the predictions of a sparse coding hypothesis in which information is represented efficiently by a small number of activated neurons.This suggests that early brain areas adopt an efficient strategy for coding information even when animals are raised in a severely limited visual environment where sensory inputs have an unnatural statistical structure.

View Article: PubMed Central - PubMed

Affiliation: Graduate School of Frontier Biosciences, Osaka University, Suita, Osaka 565-0871, Japan.

ABSTRACT
Altered sensory experience in early life often leads to remarkable adaptations so that humans and animals can make the best use of the available information in a particular environment. By restricting visual input to a limited range of orientations in young animals, this investigation shows that stimulus selectivity, e.g., the sharpness of tuning of single neurons in the primary visual cortex, is modified to match a particular environment. Specifically, neurons tuned to an experienced orientation in orientation-restricted animals show sharper orientation tuning than neurons in normal animals, whereas the opposite was true for neurons tuned to non-experienced orientations. This sharpened tuning appears to be due to elongated receptive fields. Our results demonstrate that restricted sensory experiences can sculpt the supranormal functions of single neurons tailored for a particular environment. The above findings, in addition to the minimal population response to orientations close to the experienced one, agree with the predictions of a sparse coding hypothesis in which information is represented efficiently by a small number of activated neurons. This suggests that early brain areas adopt an efficient strategy for coding information even when animals are raised in a severely limited visual environment where sensory inputs have an unnatural statistical structure.

No MeSH data available.


Orientation representation by a population of neurons in area 17.Top, Schematic illustration of orientation tuning curves of single neurons. The density of overlaid curves indicates the approximate relative number of cells (curves were removed for clarity when needed). Five curves are highlighted in colours for visual purposes. Bottom, Population orientation tunings were obtained as follows. Two-dimensional tuning surfaces were averaged across neurons. By integrating the plot radially (thereby removing the spatial frequency dimension), population tuning to orientation was obtained. The grey bands indicate the s.e.m. of the bootstrap distributions. (a) Tuning for v-goggled cats. (b) Simulated tuning. Data from normal cats were resampled so that the number of cells in each of the 3 orientation groups matched that from v-goggled cats. (c) Simulated tuning. Data from normal cats were resampled so that the number of cells tuned to the vertical and those tuned to the non-vertical orientations (i.e., horizontal + oblique as a single group) matched that from v-goggled cats. In this curve, the ratio of the number of cells for the non-vertical orientations (horizontal vs. oblique) was identical to that in normal cats. (d) Tuning for normal cats.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4644951&req=5

f6: Orientation representation by a population of neurons in area 17.Top, Schematic illustration of orientation tuning curves of single neurons. The density of overlaid curves indicates the approximate relative number of cells (curves were removed for clarity when needed). Five curves are highlighted in colours for visual purposes. Bottom, Population orientation tunings were obtained as follows. Two-dimensional tuning surfaces were averaged across neurons. By integrating the plot radially (thereby removing the spatial frequency dimension), population tuning to orientation was obtained. The grey bands indicate the s.e.m. of the bootstrap distributions. (a) Tuning for v-goggled cats. (b) Simulated tuning. Data from normal cats were resampled so that the number of cells in each of the 3 orientation groups matched that from v-goggled cats. (c) Simulated tuning. Data from normal cats were resampled so that the number of cells tuned to the vertical and those tuned to the non-vertical orientations (i.e., horizontal + oblique as a single group) matched that from v-goggled cats. In this curve, the ratio of the number of cells for the non-vertical orientations (horizontal vs. oblique) was identical to that in normal cats. (d) Tuning for normal cats.

Mentions: We next examined how a population of area 17 neurons in v-goggled cats represents orientation. Figure 6a illustrates the likely answer to this question. As shown above, cells were sharply tuned to orientation if they preferred the vertical orientation, whereas they were broadly tuned otherwise. In agreement with previous studies23456789 and our optical imaging data, a majority of neurons (102 of 147 cells) were tuned to the vertical orientation. Interestingly, the effects were not equal for the non-experienced orientations; although the oblique orientation was closer to the experienced orientation than the horizontal orientation, there were fewer neurons tuned to the oblique orientation than the horizontal orientation (17 vs. 27 cells, respectively). This difference could not be due to the residual oblique effect found in normal adult animals252627. First, these numbers for v-goggled cats were unlikely to be observed by chance when the distribution of cells in normal adult cats26 was assumed (p < 0.005, binomial test). Second, the corresponding numbers in age-matched control cats (131 vs. 54 cells, respectively) did not support the oblique effect. As a consequence, when an orientation tuning curve for a population of neurons was obtained by summing individual orientation tuning curves for recorded neurons (Fig. 6, bottom), the population tuning curve exhibited significantly lower responses to the oblique orientation (45° and 135°) than to the horizontal orientation (0°) (p < 0.001, resampling; Fig. 6a). This was not explained by the oblique effect252627 in normal cats either (Fig. 6d), because such local troughs did not emerge around the oblique orientation in the tuning curve when neurons from normal cats were resampled so that cells tuned to the oblique and horizontal orientations were selected at the same probability as in normal cats (Fig. 6c; compare with Fig. 6b). Furthermore, population orientation tuning was significantly narrower in the actual v-goggled cats than in the resampled data (Fig. 6a vs. b; p < 0.001, resampling), reflecting the sharpened orientation tuning of single neurons tuned to the vertical orientation in v-goggled cats.


Supranormal orientation selectivity of visual neurons in orientation-restricted animals.

Sasaki KS, Kimura R, Ninomiya T, Tabuchi Y, Tanaka H, Fukui M, Asada YC, Arai T, Inagaki M, Nakazono T, Baba M, Kato D, Nishimoto S, Sanada TM, Tani T, Imamura K, Tanaka S, Ohzawa I - Sci Rep (2015)

Orientation representation by a population of neurons in area 17.Top, Schematic illustration of orientation tuning curves of single neurons. The density of overlaid curves indicates the approximate relative number of cells (curves were removed for clarity when needed). Five curves are highlighted in colours for visual purposes. Bottom, Population orientation tunings were obtained as follows. Two-dimensional tuning surfaces were averaged across neurons. By integrating the plot radially (thereby removing the spatial frequency dimension), population tuning to orientation was obtained. The grey bands indicate the s.e.m. of the bootstrap distributions. (a) Tuning for v-goggled cats. (b) Simulated tuning. Data from normal cats were resampled so that the number of cells in each of the 3 orientation groups matched that from v-goggled cats. (c) Simulated tuning. Data from normal cats were resampled so that the number of cells tuned to the vertical and those tuned to the non-vertical orientations (i.e., horizontal + oblique as a single group) matched that from v-goggled cats. In this curve, the ratio of the number of cells for the non-vertical orientations (horizontal vs. oblique) was identical to that in normal cats. (d) Tuning for normal cats.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4644951&req=5

f6: Orientation representation by a population of neurons in area 17.Top, Schematic illustration of orientation tuning curves of single neurons. The density of overlaid curves indicates the approximate relative number of cells (curves were removed for clarity when needed). Five curves are highlighted in colours for visual purposes. Bottom, Population orientation tunings were obtained as follows. Two-dimensional tuning surfaces were averaged across neurons. By integrating the plot radially (thereby removing the spatial frequency dimension), population tuning to orientation was obtained. The grey bands indicate the s.e.m. of the bootstrap distributions. (a) Tuning for v-goggled cats. (b) Simulated tuning. Data from normal cats were resampled so that the number of cells in each of the 3 orientation groups matched that from v-goggled cats. (c) Simulated tuning. Data from normal cats were resampled so that the number of cells tuned to the vertical and those tuned to the non-vertical orientations (i.e., horizontal + oblique as a single group) matched that from v-goggled cats. In this curve, the ratio of the number of cells for the non-vertical orientations (horizontal vs. oblique) was identical to that in normal cats. (d) Tuning for normal cats.
Mentions: We next examined how a population of area 17 neurons in v-goggled cats represents orientation. Figure 6a illustrates the likely answer to this question. As shown above, cells were sharply tuned to orientation if they preferred the vertical orientation, whereas they were broadly tuned otherwise. In agreement with previous studies23456789 and our optical imaging data, a majority of neurons (102 of 147 cells) were tuned to the vertical orientation. Interestingly, the effects were not equal for the non-experienced orientations; although the oblique orientation was closer to the experienced orientation than the horizontal orientation, there were fewer neurons tuned to the oblique orientation than the horizontal orientation (17 vs. 27 cells, respectively). This difference could not be due to the residual oblique effect found in normal adult animals252627. First, these numbers for v-goggled cats were unlikely to be observed by chance when the distribution of cells in normal adult cats26 was assumed (p < 0.005, binomial test). Second, the corresponding numbers in age-matched control cats (131 vs. 54 cells, respectively) did not support the oblique effect. As a consequence, when an orientation tuning curve for a population of neurons was obtained by summing individual orientation tuning curves for recorded neurons (Fig. 6, bottom), the population tuning curve exhibited significantly lower responses to the oblique orientation (45° and 135°) than to the horizontal orientation (0°) (p < 0.001, resampling; Fig. 6a). This was not explained by the oblique effect252627 in normal cats either (Fig. 6d), because such local troughs did not emerge around the oblique orientation in the tuning curve when neurons from normal cats were resampled so that cells tuned to the oblique and horizontal orientations were selected at the same probability as in normal cats (Fig. 6c; compare with Fig. 6b). Furthermore, population orientation tuning was significantly narrower in the actual v-goggled cats than in the resampled data (Fig. 6a vs. b; p < 0.001, resampling), reflecting the sharpened orientation tuning of single neurons tuned to the vertical orientation in v-goggled cats.

Bottom Line: Our results demonstrate that restricted sensory experiences can sculpt the supranormal functions of single neurons tailored for a particular environment.The above findings, in addition to the minimal population response to orientations close to the experienced one, agree with the predictions of a sparse coding hypothesis in which information is represented efficiently by a small number of activated neurons.This suggests that early brain areas adopt an efficient strategy for coding information even when animals are raised in a severely limited visual environment where sensory inputs have an unnatural statistical structure.

View Article: PubMed Central - PubMed

Affiliation: Graduate School of Frontier Biosciences, Osaka University, Suita, Osaka 565-0871, Japan.

ABSTRACT
Altered sensory experience in early life often leads to remarkable adaptations so that humans and animals can make the best use of the available information in a particular environment. By restricting visual input to a limited range of orientations in young animals, this investigation shows that stimulus selectivity, e.g., the sharpness of tuning of single neurons in the primary visual cortex, is modified to match a particular environment. Specifically, neurons tuned to an experienced orientation in orientation-restricted animals show sharper orientation tuning than neurons in normal animals, whereas the opposite was true for neurons tuned to non-experienced orientations. This sharpened tuning appears to be due to elongated receptive fields. Our results demonstrate that restricted sensory experiences can sculpt the supranormal functions of single neurons tailored for a particular environment. The above findings, in addition to the minimal population response to orientations close to the experienced one, agree with the predictions of a sparse coding hypothesis in which information is represented efficiently by a small number of activated neurons. This suggests that early brain areas adopt an efficient strategy for coding information even when animals are raised in a severely limited visual environment where sensory inputs have an unnatural statistical structure.

No MeSH data available.