Why is there this extra term for this correlation function?

[AndrewGRQTF]

Let's say we have a Dirac field $\Psi$ and a scalar field $\varphi$ and we want to compute this correlation function $$<0|T \Psi _\alpha (x) \Psi _\beta (y) \varphi (z_1) \varphi (z_2)|0>$$ $$= \frac {1}{i} \frac{\delta}{\delta \overline{\eta}_\alpha(x)} i \frac{\delta}{\delta \eta_\beta(y)} \frac {1}{i} \frac{\delta}{\delta J(z_1)} \frac {1}{i} \frac{\delta}{\delta J(z_2)} iW(\overline{\eta}, \eta, J) |_{\overline{\eta}, \eta, J = 0}$$ $$= c \int d^4 w_1 d^4 w_2 [S(x-w_1)S(w_1 - w_2) S(w_2 -y)]_{\alpha \beta} \Delta (z_1 - w_1) \Delta (z_2 - w_2) + \mathrm{the\ previous\ term\ with \ z_1\ and\ z_2\ exchanged}$$
where c is a constant

My question is: why do we have the last term in the equation? Why do we include the term with $z_1$ and $z_2$ exchanged?

 

[AndrewGRQTF]

Here is additional information