Testing engraving. The test pool: 2 dwarves Legendary +5 Engravers, 2 Dwarves 0 skill engravers. 2 dwarves 500 agility, creativity, spacial, and kinaesthetic, and 2 with 1,500 attributes. Mixed and matched, of course. All dwarves were given a 10x10 plot to smooth and then engrave.
Subject 1: Legendary +5 - Att 1500
100 smoothings = +1 Agility, +1 Kin
100 engravings = +1 Creativity, +1 Spacial
Total +1 to all stats hitting 1,501 in each, plus 21 masterpiece engravings.
Subject 2: Legendary +5 - Att 500
100 Smoothings = Exact same growth.
100 Engravings = Exact same again
Total Identical to the Attribute 1,500 dwarf. 25 Masterpieces.
Subject 3: Zero skill - Att 500
100 Smoothings = +13 Agi, +12 Cre, +12 Spatial, +13 Kin
100 Engravings =+21 Agi, +21 Cre, +21 Spatial, +21 Kin
NOTE: It is important to take note that this dwarf had the most problems, namely food, breaks, booze, and other such dwarven things, and as such was the slowest to perform his job, by far.
Subject 4: Zero skill - Att 1500
100 Smoothings = +14 Agi, Cre +13, +11 Spacial, +14 Kin
100 Engravings = +13 Agi, +13 Cre, +12 Spacial, +13 Kin
DFHack was used to create 4 blocks of 10x10x1 obsidian, mined out by 2 Legendary +5 miners. The Legendary +5 engravers were allowed to do their jobs. The no-skill engravers had to be burrowed, so they could both work at the same time and cut down on overall work time. Runesmith was used to adjust skill levels and attribute numbers, and to check attribute progress.
It seems extremely obvious to me that the time spent performing a job determines how much attribute gain that job gives. Cluttering you workshops should cause these workshops to give higher attribute returns, in case you're wanting to train up strength via masonry, it may pay off to overload the mason's shop with clutter to drag out the length of each task.