Treating ordinal independent values in linear regression

Hello!

I wish to build a linear model using > 40 independent variables. Of
these, 21 variables are ordinal and each takes values from 1 to 5.
Also, each variable has a peaked distribution at 3 (being the middle
value).

Could anyone shed light on how such variables should be treated to
build the regression model? Would it be correct to use them as they
are... or convert to dummy variables? Importantly, how would one
explain exclusion/inclusion of select dummy variables (for each ordinal
variable) when using stepwise regression?

Many thanks in advance for your replies...
Rahu

.