Can I use binary covariate in anova? I am performaning an anova and have resilience and anxiety scores that want to use as covariates. These are coded into high and low. 
If I imput them as covariates I obtain significant results. However if I imput them as between subjects factors I dont get any significant results.
Can I add them to the model as covariates? 
Thank you
 A: First, as always, why are you dichotomizing? Are your high/low cutoffs theoretically backed (e.g. your cutoff for high anxiety has clinical relevance) or is it a median split or other arbitrary cutoff? If the latter, just use the continuous data. You are throwing away valuable information. The power problem gung mentioned is likely not a problem though. Dichotomizing affects the power to detect an effect of the dichotomized variable. Since you use the term covariate, I assume there is a third variable, X, that you are really interested in. Dichotomizing anxiety, etc. does not affect the power for tests of X. If you are interested in the effects of anxiety and resilience, then you will have diminished power.
Yes, you can add them as covariates, depending on what your research questions are. If all you are interested in is the affect of X on Y and want to control for baseline levels of anxiety and resilience, then including them in your model as covariates is the correct model. I will add a note of caution. There is a long literature in clinical psychology (and elsewhere, but you seem to be doing clinical psych work) criticizing how ANCOVA models are used and interpreted. Probably the best example is Miller, G. A., & Chapman, J. P. (2001). Misunderstanding analysis of covariance. Journal of Abnormal Psychology, 110(1).
Second, your description of how you "input" the variables tells me you are using SPSS, and may not know the difference between the models SPSS uses when you put a variable in the covariate box vs. the factor box. When X, anxiety and resilience put in the factor box, the model includes interactions between all factor variables by default. When anxiety and resilience are put in the  covariate box, and X in the factor box, no interactions are included by default. So as covariates the model is
$$
y = \beta_0 + \beta_1 ~ X + \beta_2 ~ \textrm{ANXIETY} + \beta_3 ~ \textrm{RESILIENCE} + \epsilon
$$
As factors the model is
$$
y = \beta_0 + \beta_1 ~ X + \beta_2 ~ \textrm{ANXIETY} + \beta_3 ~ \textrm{RESILIENCE} + \beta_4 ~ X ~ \textrm{ANXIETY} + \beta_5 ~ X ~ \textrm{RESILIENCE} + \beta_6 ~ \textrm{ANXIETY} ~ \textrm{RESILIENCE} + \beta_7 ~ X ~ \textrm{ANXIETY} ~ \textrm{RESILIENCE} + \epsilon
$$
Very different models, that need to be interpreted very differently.
