I think this is partly a matter of definition: what sorts of processes are we willing to call "a computation"?
Let's suppose we're charged with the task to compute, given any natural number $n$, the $n$th Fibonacci number.
It is defined as follows: if $F(n)$ stands for the $n$th Fibonacci number, then $F(1) = F(2) = 1$ and $F(n) = F(n-1) + F(n-2)$ for all $n > 2$.
We could actually define it that way in most programming languages.
For instance, in C:
int fibonacci(int n)
{
if (i < 2)
{
return 1;
}
else
{
return fibonacci(n-1) + fibonacci(n-2);
}
}
This is a valid (and inefficient) way to compute Fibonacci numbers in C.
But does it define a computation to compute the $n$th Fibonacci number? You tell me.
In C, the order of evaluation of arguments of + is undefined. One implementation may always evaluate the left operand first, then the right. Another may choose an evaluation order depending on the weather. Yet another may evaluate them in parallel (I do believe this is valid). C programs mostly run in multiprocessing environments, so a parallel implementation could actually be implemented most of the time.
Let's suppose you use such a parallel implementation. It is certainly a way to compute the $n$th Fibonacci number. But is it "a computation"? I'm not sure. I think for most of us, a computation is a stepwise process, executed by a single processor strictly sequentially, one step at a time. Not because computation must be done that way (that isn't the case) but because we don't call the process "a computation" otherwise.
Modern computers and the human brain don't quite work strictly sequentially. It's our conception of computation, an abstraction of what really happens, that does. We say a computer or a human brain performs computation when it does something that is equivalent to what that abstraction does - even when how it actually does it deviates from that.
This is part of why the human brain finds it difficult to come to terms with inherently concurrent or parallel computational processes. We're trained to think of computation as a strictly sequential single-processor process.
If you want to meet a different perspective on computation, familiarize yourself with dataflow processing, e.g. Unix pipes, or typical setups for signal processing, such as audio and video processing. What those systems do is computation, and it is a pretty mainstream way of using computers, but it is not adequately described by Turing machines.
