What test do I use in order to analyze a within participants repeated measure experiment? @RyanB thank you for your quick reply and your help!
I conducted an experiment supporting that men are attracted by women in red. I recruited 30 men and 25 women who followed exactly the same procedure: They had to listen to 4 dialogues between a man and a woman which one of them lasts for 30-40 sec. They hear the first dialogue and then I ask them if they believe that the man is attracted to woman, their answer shall be yes/no. Then I show them to photos, a woman with green shirt and a woman with a red one and ask them to choose the photo which, they think, depicts the woman of the dialogue they just heard. Then they listen to the second dialogue and they do exactly the same until they finish. (1st dialogue- attracted/not attracted- red/green). 
The reason why I do the same experiment to women (they have to rate the attractiveness of other women instead of men) is to show the difference between the color preferences between men and women and to replicate one part of the previous experiment which showed exactly the same. I used three different girls, and I randomized my data (dialogues, colors, girls) in excel.
So, if A is red and B is green, I've got A1, A2 and A3 and B1, B2 and B3. And my combinations might be A1-B2, B3-A2, A3-B2, B1-A2 etc.
 A: Create a table that crosstabulates Perceived Attraction (“yes/no”) x Color Choice x Gender x Trial. This should give you a good sense of whether there are main effects or interactions of gender and trial number on color choice or perceived attraction, and you will want to include it in your write-up. 
Suppose you are primarily interested in color choice as opposed to perceived attraction. A basic approach that takes into account the dependence within participants is to fit a binomial regression of color choice on gender, trial number, and their interaction, and include a participant-level random effect. You can do this in R and test the fixed effects using: 
 fm1 <- lmer(Color ~ Female * Trial + (1 | ID), data=dat, family=binomial(), REML=F)
 fm2 <- lmer(Color ~ Female + Trial + (1 | ID), data=dat, family=binomial(), REML=F)
 fm3 <- lmer(Color ~ Female + (1 | ID), data=dat, family=binomial(), REML=F)
 fm4 <- lmer(Color ~ Trial + (1 | ID), data=dat, family=binomial(), REML=F)
 anova(fm1, fm2) # Tests for interaction
 anova(fm2, fm3) # Tests for main effect of trial
 anova(fm2, fm4) # Tests for main effect of gender

where dat is your data matrix and all variables are coded as factors. Report the results as you would an ANOVA except that your test statistics will be $\chi^2$ instead of $F$. Test your random effect term for participants:
 fm5 <- lmer(Color ~ Female * Trial + (1 | ID), data=dat, family=binomial(), REML=T)
 fm6 <- lm(Color ~ Female * Trial, data=dat, family=binomial())
 pchsqi(as.numeric(2*(logLik(fm5) - logLik(fm6)), 1, lower=F) #A conservative test

Suppose your interaction and the random effect term is significant. Report and interpret the regression output fm5. For example, males might 4 times more likely than females to choose red, and this effect might be strongest for trial 4 vs. 1, etc. 
You might further hypothesize that the color preference effect only exists or is strongest for males who were successfully primed by the attraction dialogue. To test such a hypothesis, include perceived attraction and possibly perceived attraction in the previous trial as covariates and reanalyze. 
Finally, it is possible that there are effects of your stimuli (dialogue, photos, color-to-photo assignment, orderings of the combination of the aforementioned variables), but I doubt you have enough data to analyze these, so I would just carefully report how you randomized them (I’m assuming you assigned orderings randomly to participants).
