If there is no "using" clause, then gnuplot looks at each line to
decide if there is valid data, and if so how many columns of data.
Gnuplot has historically treated each line of input as if it were
independent of any others. That means that if different lines in the
data file have different numbers of columns, or different numbers of
valid columns, you get rather odd results.
So "using 1:2" will accept only input lines that indeed contain valid
data in the first two columns. If there is no "using" clause, then
lines containing 0 or 1 valid columns will be accepted without any
error message but may well not be plotted as you expected.
Note:
This will _change_ in gnuplot version 5 whenever that is released.
It has already been changed in the development version so that in the
absence of a "using" clause the program will effectively construct one
based on the first line of data. That means if later input lines contain
fewer valid columns than the first line, you'll get an error message.
> Karl