You are making the same assumption and the same mistake
here as many have
before you... so don't feel that you are alone in
struggling with the
concepts. Pixels per inch (e.g. 72x72 or 300x300) are NOT
the same thing as
dots per inch at the printer (4800x2400.) In printing the
number of dots
required to make up one image pixel may range from none
(for the colour
white - or really that is representing parts of the
printed document with an
absence of any ink being laid down by the printer) to the
other extreme of
several dots of each of the available inks to represent a
single pixel.

When you ONLY make the change in resolution (72ppi to
300ppi) in the editor
all you actually do is reduce the "apparent" native size
of a 100% scaled
image at the print stage.

Thus an image of 1600x1800 @72ppi image will tell you (in
the image
properties in the editor) that it is ca 22x25 (printed)
inches when scaled
at 100%, BUT, the exact same 1600x1800 image @300ppi will
tell you that it
is 5.3x6 (printed) inches when scaled at 100% - AND there
is no re-sampling
done to get there. It is merely a scaling value for
printing purposes.
Further, you can also set your printer to print that image
at various
resolutions in dots per inch - whereupon the PRINTER
DRIVER makes changes by
interpolation to the image file between receiving it in
the print queue and
transmitting the data to the actual print heads - this is
something you have
only very limited control over and only by changing print
quality parameters
in the printer dialog between, say, draft quality and
photo quality outputs
and altering (usually) the selected paper type to
accommodate the print
quality (e.g. plain paper or photo paper). The printer, in
turn, computes
the way that each pixel is managed as far as the inks, and
quantity of inks,
laid down on the actual paper output.

So far I have not talked about making changes, except for
the printer
quality settings, that affect image quality at the print
stage... to do that
would involve a re-sampling step such as in the image
editor. Taking the
original 72ppi image, for example, and command it to be
300ppi (ONLY reduces
"apparent" print output size) BUT also command a change in
image size -
THEN - you will have made a change that is qualitative vs.
the original
image file and that change will have required resampling
(probably, and
almost certainly, creating data that was not part of the
image in the first
place) in the editor stage and before the printer gets it.

Using the same image I used above: 1600x1800 image @300ppi
the image
properties will tell you that it is 5.3x6 (printed) inches
when scaled at
100% - I could command this to remain at 300ppi resolution
but change the
print size to, say, 10x12 (a multiplier of 4x total pixels
and a dimension
multiplier in each direction of 2x) and that image will
then have apparent
values of ca 3200x3600 @300ppi - the extra pixels are
created out of some
(usually selectable) resampling algorithm - but they are
nevertheless
machine created and incorporated into the original image.
The native print
size, scaled at 100% in the printer, is then as selected,
10x12inches
@300pixels per inch.

Journalist

That is a very excellent explanation of the issue, however I
don't understand it in a practical way. To use your
example, is it best to resample an image to 3200x3600 @
300ppi if you wish to make a 10x12 print, or do the photo
apps handle this in the background before they send the data
to the printer?