How to calculate the amount of standards deviations a measure is from another value I have a theoretical value $\gamma_t$ and a measured values $\gamma_m$, with uncertainties $\sigma_{\gamma_t}$ and $\sigma_{\gamma_m}$ respectively. If the errorbars of $\gamma_m$ is inside of $\gamma_t$-'s, as illustrate below, would that mean that $\gamma_m$ is within one standard deviation of $\gamma_t$? Or would the best guess for $\gamma_m$ have to be with the errorbar for $\gamma_t$?
A specific example:
$$ \gamma_t = 4.9\pm 1.1, \quad \gamma_m = 3.1\pm 1.5$$
In this case, the errorbars overlap but $\gamma_m=3.1$ does not lie within one standard deviation of $\gamma_t$. Would $\gamma_t$ be within one $\sigma_{\gamma_t}$ or would it instead be $n$ away, where $n$ is:
$$ \gamma_t = n\sigma_{\gamma_t}+\gamma_m $$
$$ \Rightarrow n \approx 1.6 $$
I hope I made my case clear. Here is the illustration:

 A: When you want to see if two quantities with uncertainty agree, you should compare their difference with zero.  If $\gamma_t-\gamma_m=0$ within the uncertainty, they agree.  The difference is the best way to check the degree of (dis)agreement, too.  The number of standard deviations, $n_\sigma$, is:
$$n_\sigma = \frac{\gamma_t-\gamma_m}{\sigma},$$
where $\sigma$ is the total uncertainty on the difference.
Assuming the two quantities are independent, then their uncertainties add in quadrature:
$$ \sigma = \sqrt{{\sigma_t}^2 + {\sigma_m}^2}.$$
If the two quantities are not independent, a more detailed error analysis is needed.  For example if both $\gamma_t$ and $\gamma_m$ were calculated using the length of some part of the experimental apparatus, $\ell \pm \sigma_\ell$, the uncertainty in that length would contribute to both $\sigma_t$ and $\sigma_m$.  In that case the two are not independent.
Assuming your two quantities are independent,
$$ \gamma_t - \gamma_m = (4.9 \pm 1.1) - (3.1\pm 1.5) = 1.8 \pm 1.9,$$
which does agree with zero within the combined "$1\sigma$" uncertainty.  In your case $n_\sigma<1$.
