expectation from temporal sequences influences ... - Adrien Chopin

0.7. 1. 11. 12. 111. 121. 112. 122. 1111. 1212. 1121. 1222. 2. 22. 21. 222. 212. 221. 211. 2222. 2121. 2212. 2111 p(1). Difference between numbers of 2s and 1s.
2MB taille 3 téléchargements 597 vues
1

Adrien Chopin 2 Madison Capps 1 Pascal Mamassian

EXPECTATION FROM TEMPORAL SEQUENCES INFLUENCES BINOCULAR RIVALRY

1 : Université Paris Descartes, CNRS ; 2 : Massachussetts Institute of Technology

Corresponding author: [email protected]

INTRODUCTION

RESULTS

We investigate here the implicit encoding of a temporal sequence of visual events and the expectation to complete the sequence. For this purpose, we tested the extent to which a series of non-rivalrous patterns can influence the dominant perception in binocular rivalry. We rely on and extend the pattern suppression phenomenon: when rivalrous oriented Gabors follow non-rivalrous Gabors, observers usually perceive the repeated orientation less often (Brascamp, Knapen, Kanai, van Ee & van den Berg, 2007).

Observer's dominant percept was deduced from their spatial frequency response.

0.6

p(1)

112 1121 121 2111 1 211 11

0.5 0.4

1111

0.3

111

-4

-2

p(1)

112

0.5 0.4

1111

0.3

111

-4

1121 121 2111 1 211 11 -2

LLL

LLLL

RLLL

RLL

LLRL

LRL

RLRL

+R

RRL

RRRR LRRR

LRR

RLR

RRLR LRLR

LLR

111

0.43

1111

2111

211

1121

4

221

121

2121

p(L)

2222

222

N=8

= 64 measures by observer

2

4

0.38

Another factor was related to alternation: after the series 2121, orientation 2 was more often perceived than predicted by our model. Furthermore, a model with one parameter for the alternation component, one for the gambler's fallacy component and one for the last item adaptation was better than a model with three parameters for adaptation of the three last items (log(likelihood) = -3784 vs -3785).

221

= 64 measures by observer

DISCUSSION

ALTERNATION COMPONENT

0

N=8

0.25

%1

21

%R RRR

0.29

be for e

be for e

for e

be +L

+1

11

%L

0.49

e for

LR

0.5

A strong factor was related to adaptation: if the last item was orientation 1, then orientation 2 was more likely to be perceived in the rivalrous stage. Nevertheless, the combination of the gambler's fallacy and adaptation hypotheses predicted better the data than adaptation alone. A model with one parameter for the last item adaptation and one parameter for the gambler's fallacy component was better than a model with two parameters for adaptation of the two last items (log(likelihood) = -3785 vs -3787).

be

RR

1

+2

e for

e for

RL

0.56

be

be

LL

R

+L

+R

0.42

2121 21

22 1222 2212

122 2 212

Difference between numbers of 2s and 1s

Combined to

L

= 64 measures by observer

2

12 1212

0.75

N=8

Difference between numbers of 2s and 1s

0.6

- Series are non-rivalrous Gabors oriented either to the Response? left (L) or to the right (R). Is spatial frequency - Series length varies from 1 to 4 higher or lower? items, for instance LLRL. - Each series is followed by a pair of rivalrous Gabors. - The spatial frequency of the Gabors when rivalrous is different from that of the Gabors within a series. - Task : is the seen spatial frequency during the rivalrous stage higher or lower than the one seen during the series?

22 1222 2212

2121 221 21

0

ADAPTATION COMPONENT

time

12 122 1212 2 212

2222

= 32 measures by observer

111 1 111 11 211 1 112 1 121 1 211 112 212 1 21 12 121 2 221 122 2 212 221 2 122 2 22 222 222 2

LR

222

N=8

0.25

p(1)

0.7

0.5

LLL L LLL LL RLL LLRL L LRL L RLL LLR RLR L RL LR LRL R RRL LRR R RLR RRL LRRR R RR RRR RRR R

GAMBLER'S FALLACY COMPONENT

An example of

SERIES

0.75

The dominant percept during the rivalrous stage was largely predictable from the current series. If orientation 1 was shown more often than orientation 2 during the series, then orientation 2 was seen more often in the rivalrous stage. This suggests that the visual system tries to complete the series by equalling the number of orientations 1 and 2. This behavior is refered as the gambler's fallacy component.

METHODS

a series:

Model : Gambler's fallacy + Adaptation + Alternation

These results are partially consistent with the phenomenon of pattern completion discovered with the ambiguous motion quartet (Maloney, Martello, Sahm & Spillmann, 2005). In conclusion, binocular rivalry is not only influenced by adaptation of the perceived orientations, but also by more complex temporal structures.

CONCLUSION

- Expectancies play a role in binocular rivalry (by the gambler's fallacy and alternation components). - Since attention was drawn to spatial frequency instead of orientation, expectancies effects reflect more the implicit predictions of the visual system than some strategy of the observer.

REFERENCES

Brascamp, J. W., Knapen, T. H. J., Kanai, R., van Ee, R., & van den Berg, A. V. (2007). Flash suppression and flash facilitation in binocular rivalry. Journal of Vision, 7(12), 1-12. doi:10.1167/7.12.12 Maloney, L. T., Martello, M. F. D., Sahm, C., & Spillmann, L. (2005). Past trials influence perception of ambiguous motion quartets through pattern completion. Proceedings of the National Academy of Sciences, 102(8), 3164-3169.

ACKNOWLEDGEMENTS

This work was supported by funds from the French Research Ministry and the CNRS. Impression reprographie Université Paris Descartes.