Check normality assumptions.

The differences don’t seem to follow a normal distribution.

But for n=16, there aren’t any outliers nor is the skewness particularly strong, so we can safely use the t-procedures.