Here are two proofs using only the angle sum identities, the fact the trig functions are periodic, and their values at $0$. It is inspired on the connection to rotations and the fact that rotations don't change the sizes of things, but do not actually assume that connection.
Define the matrix
$$ A(\theta) = \left( \begin{matrix} \cos(\theta) & \sin(\theta) \\ -\sin(\theta) & \cos(\theta) \end{matrix} \right) $$
By the angle addition formula, we see that $A(\theta + \varphi) = A(\theta) A(\varphi)$; in particular, $A(n \theta) = A(\theta)^n$.
There are arbitrarily large integer multiples of $\theta$ that are arbitrarily close to integer multiples of $2 \pi$, which gives
$$ A(\theta)^n = A(n \theta) \approx I $$
where $I$ is the identity matrix. By taking determinants, we get
$$ (\cos(\theta)^2 + \sin(\theta)^2)^n \approx 1$$
for arbitrarily large $n$; since the determinant is a real nonnegative number, the only possibility is that
$$ \cos(\theta)^2 + \sin(\theta)^2 = 1$$
To get the details right, we select sequences of integers $a_n, b_n$ with
$$ \lim_{n \to \infty} a_n \theta - 2 \pi b_n = 0 $$
$$ \lim_{n \to \infty} a_n = +\infty $$
e.g. this can be done by continued fractions. Then,
$$ \lim_{n \to \infty} A(a_n \theta) =
\lim_{n \to \infty} A(a_n \theta - 2 \pi b_n)
= A(0) = I$$
and consequently
$$ \lim_{n \to \infty} (\det A(\theta))^{a_n} = 1 $$
Another method in the same vein is to also define vectors
$$ v(\theta) = \left( \begin{matrix} \sin(\theta) \\ \cos(\theta) \end{matrix} \right) $$
so that $A(\theta) v(\varphi) = v(\theta + \varphi)$
However, we know that for every angle $\theta$:
$$\frac{1}{2} \leq \max(\sin(\theta)^2, \cos(\theta)^2) \leq 1 $$
In particular, we have
$$ \frac{1}{2} \leq \| v(\theta) \| \leq 2 $$
and furthermore, the $v(\theta)$ span $\mathbb{R}^2$.
By comparing the lengths of $v(\varphi)$ and $A(\theta) v(\varphi)$, we know that all of the eigenvalues of $A(\theta)$ must have magnitude lying between $1/2$ and $2$.
However, this remains true for $A(\theta)^n$, and consequently, all of the eigenvalues of $A(\theta)$ must be roots of unity, and thus the determinant is either $1$ or $-1$, and it can't be $-1$.
This information helps others identify where you have difficulties and helps them write answers appropriate to your experience level.I have no difficulties. This is not homework. I want different levels of answering because I want to know how different mathematicians prove this in different ways. This community will always be regarded as a place for little children to cheat on their homework if all of you keep assuming that every question that sort of looks like homework is homework without even enquiring! – Nick Dec 14 '13 at 23:47