Limits...
Contextual Congruency Effect in Natural Scene Categorization: Different Strategies in Humans and Monkeys (Macaca mulatta).

Collet AC, Fize D, VanRullen R - PLoS ONE (2015)

Bottom Line: In real conditions, objects (either animate or inanimate) are never isolated but embedded in a complex background made of multiple elements.We designed a rapid visual categorization task, Animal versus Non-Animal, using as contexts both real scenes photographs and noisy backgrounds built from the amplitude spectrum of real scenes but with randomized phase spectrum.We showed that even if the contextual congruency effect was comparable in both species when the context was a real scene, it differed when the foreground object was surrounded by a noisy background: in monkeys we found a similar congruency effect in both conditions, but in humans the congruency effect was absent (or even reversed) when the context was a noisy background.

View Article: PubMed Central - PubMed

Affiliation: Université de Toulouse, UPS, Centre de Recherche Cerveau et Cognition, Toulouse, France; CNRS, CerCo, Toulouse, France.

ABSTRACT
Rapid visual categorization is a crucial ability for survival of many animal species, including monkeys and humans. In real conditions, objects (either animate or inanimate) are never isolated but embedded in a complex background made of multiple elements. It has been shown in humans and monkeys that the contextual background can either enhance or impair object categorization, depending on context/object congruency (for example, an animal in a natural vs. man-made environment). Moreover, a scene is not only a collection of objects; it also has global physical features (i.e phase and amplitude of Fourier spatial frequencies) which help define its gist. In our experiment, we aimed to explore and compare the contribution of the amplitude spectrum of scenes in the context-object congruency effect in monkeys and humans. We designed a rapid visual categorization task, Animal versus Non-Animal, using as contexts both real scenes photographs and noisy backgrounds built from the amplitude spectrum of real scenes but with randomized phase spectrum. We showed that even if the contextual congruency effect was comparable in both species when the context was a real scene, it differed when the foreground object was surrounded by a noisy background: in monkeys we found a similar congruency effect in both conditions, but in humans the congruency effect was absent (or even reversed) when the context was a noisy background.

No MeSH data available.


Related in: MedlinePlus

Cumulated d’ as a function of time in the “Noisy Background” condition (bins of 30ms in monkeys, quantiles of relative time in humans: mean quantile duration = 20ms SEM +/0.6ms, first quantile starts at 370ms SEM +/- 11ms).The thick red and green lines represent 0% and 100% congruency levels (respectively), while intermediate levels are depicted with thin lines. The horizontal black line indicates the significance of the difference between extreme congruency level trials (permutation test, 1000 permutations, α<0.01). Inserts on top left represent linear regressions of total cumulated d’ as a function of congruency level. Blue dots are observed values and black lines are the linear regression. Slopes, R2 and p-values are respectively sDy = 0.026, R2Dy = 0.45 pDy = 0.024, sRx = 0.054, R2Rx = 0.72, pRx = 0.001, sH = -0.027, R2H = 0.58, pH = 0.006.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4514779&req=5

pone.0133721.g004: Cumulated d’ as a function of time in the “Noisy Background” condition (bins of 30ms in monkeys, quantiles of relative time in humans: mean quantile duration = 20ms SEM +/0.6ms, first quantile starts at 370ms SEM +/- 11ms).The thick red and green lines represent 0% and 100% congruency levels (respectively), while intermediate levels are depicted with thin lines. The horizontal black line indicates the significance of the difference between extreme congruency level trials (permutation test, 1000 permutations, α<0.01). Inserts on top left represent linear regressions of total cumulated d’ as a function of congruency level. Blue dots are observed values and black lines are the linear regression. Slopes, R2 and p-values are respectively sDy = 0.026, R2Dy = 0.45 pDy = 0.024, sRx = 0.054, R2Rx = 0.72, pRx = 0.001, sH = -0.027, R2H = 0.58, pH = 0.006.

Mentions: We finally applied the d’ formula to each bin or quantile and plotted the cumulated d’ curves as shown in Figs 3 and 4.


Contextual Congruency Effect in Natural Scene Categorization: Different Strategies in Humans and Monkeys (Macaca mulatta).

Collet AC, Fize D, VanRullen R - PLoS ONE (2015)

Cumulated d’ as a function of time in the “Noisy Background” condition (bins of 30ms in monkeys, quantiles of relative time in humans: mean quantile duration = 20ms SEM +/0.6ms, first quantile starts at 370ms SEM +/- 11ms).The thick red and green lines represent 0% and 100% congruency levels (respectively), while intermediate levels are depicted with thin lines. The horizontal black line indicates the significance of the difference between extreme congruency level trials (permutation test, 1000 permutations, α<0.01). Inserts on top left represent linear regressions of total cumulated d’ as a function of congruency level. Blue dots are observed values and black lines are the linear regression. Slopes, R2 and p-values are respectively sDy = 0.026, R2Dy = 0.45 pDy = 0.024, sRx = 0.054, R2Rx = 0.72, pRx = 0.001, sH = -0.027, R2H = 0.58, pH = 0.006.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4514779&req=5

pone.0133721.g004: Cumulated d’ as a function of time in the “Noisy Background” condition (bins of 30ms in monkeys, quantiles of relative time in humans: mean quantile duration = 20ms SEM +/0.6ms, first quantile starts at 370ms SEM +/- 11ms).The thick red and green lines represent 0% and 100% congruency levels (respectively), while intermediate levels are depicted with thin lines. The horizontal black line indicates the significance of the difference between extreme congruency level trials (permutation test, 1000 permutations, α<0.01). Inserts on top left represent linear regressions of total cumulated d’ as a function of congruency level. Blue dots are observed values and black lines are the linear regression. Slopes, R2 and p-values are respectively sDy = 0.026, R2Dy = 0.45 pDy = 0.024, sRx = 0.054, R2Rx = 0.72, pRx = 0.001, sH = -0.027, R2H = 0.58, pH = 0.006.
Mentions: We finally applied the d’ formula to each bin or quantile and plotted the cumulated d’ curves as shown in Figs 3 and 4.

Bottom Line: In real conditions, objects (either animate or inanimate) are never isolated but embedded in a complex background made of multiple elements.We designed a rapid visual categorization task, Animal versus Non-Animal, using as contexts both real scenes photographs and noisy backgrounds built from the amplitude spectrum of real scenes but with randomized phase spectrum.We showed that even if the contextual congruency effect was comparable in both species when the context was a real scene, it differed when the foreground object was surrounded by a noisy background: in monkeys we found a similar congruency effect in both conditions, but in humans the congruency effect was absent (or even reversed) when the context was a noisy background.

View Article: PubMed Central - PubMed

Affiliation: Université de Toulouse, UPS, Centre de Recherche Cerveau et Cognition, Toulouse, France; CNRS, CerCo, Toulouse, France.

ABSTRACT
Rapid visual categorization is a crucial ability for survival of many animal species, including monkeys and humans. In real conditions, objects (either animate or inanimate) are never isolated but embedded in a complex background made of multiple elements. It has been shown in humans and monkeys that the contextual background can either enhance or impair object categorization, depending on context/object congruency (for example, an animal in a natural vs. man-made environment). Moreover, a scene is not only a collection of objects; it also has global physical features (i.e phase and amplitude of Fourier spatial frequencies) which help define its gist. In our experiment, we aimed to explore and compare the contribution of the amplitude spectrum of scenes in the context-object congruency effect in monkeys and humans. We designed a rapid visual categorization task, Animal versus Non-Animal, using as contexts both real scenes photographs and noisy backgrounds built from the amplitude spectrum of real scenes but with randomized phase spectrum. We showed that even if the contextual congruency effect was comparable in both species when the context was a real scene, it differed when the foreground object was surrounded by a noisy background: in monkeys we found a similar congruency effect in both conditions, but in humans the congruency effect was absent (or even reversed) when the context was a noisy background.

No MeSH data available.


Related in: MedlinePlus