The Screw Drive
Why the driver jumps out of the screw head, and why the man who fixed it in 1908 never sold it in America.

Peter Lymburner Robertson was demonstrating tools for a Philadelphia firm on a street in downtown Montreal when the blade of his screwdriver came out of the slot and opened his hand. Anyone who has driven a screw knows the movement without the blood. The tip skates sideways off the head, tears a bright arc across the wood, and leaves you holding a tool that has just done the opposite of what you asked of it. Canada’s national museum of science and technology keeps the anecdote because the cut produced a patent. Robertson went away and punched a square hole in the screw.
What he was fighting is geometrical. A slotted screw carries a straight kerf sawn clear across the head, open at both ends and no deeper in the middle than at the rim, so there is nothing in it to stop a blade travelling out sideways or riding up and away. Any recess with sloping walls suffers a quieter version of the same fault. Press a tapered driver into a tapered socket and turn it, and some fraction of the twist you apply is converted into a force wedging the two apart along the shank, which pushes the tool back at you at the exact moment you are leaning on it hardest. The trade had a name for this before the workshops did. An American Screw Company patent filed at the end of 1942 calls it throw-out: “Axial thrust endeavoring to force a screw driver out of the screw recess as a resultant of the driving torque.” Modern usage calls it cam-out. A wall that stands parallel to the screw’s axis has nowhere to send that thrust, and that single sentence is the entire case for a square hole.
Robertson opened his works at Milton, Ontario, in 1908, and the difficulty was all in the making: the socket had to be punched cold into the head in a single blow without splitting it open. His advertising card ran the whole history of the fastener across five inches, from an Archimedean screw marked B.C. to his own, under the claim that it drives like lightning and does not wobble, slip or burr. Ford’s Windsor plant was already buying, at a saving Ingenium puts near $2.60 a car, and Detroit came asking. The executives wanted a say in where the screws would be made. Robertson, who had already been badly used once by licensees overseas, declined and went home, and the square drive stayed a Canadian habit.
Ford went cruciform instead, and the received story of what he bought is wrong in two places. Henry Phillips did not invent the Phillips screw, and the cross in the head was not a new idea: John Frearson had patented a screw with a cruciform orifice in 1884. The design that carried was filed in 1932 by John P. Thompson of Portland, Oregon, whom the city directory records under “furnished rooms” and “laborer” and whom the Sunday Oregonian later described as an auto mechanic. His two patents, one for the screw and one for the driver, were granted in 1933, and the rights went by assignment to Henry F. Phillips, the managing director of a copper-mining outfit who understood what a licence was worth. Phillips built a company around them in Portland, talked the American Screw Company into manufacturing, and got the screws into Cadillacs in 1937. Thompson died in Portland in 1940. The name on the handle in your drawer belongs to the man who sold it.
The other thing everyone knows about the Phillips head is that it was designed to slip, a deliberate torque limiter for assembly lines whose clutches could not be trusted. Read what the inventors filed.
IHeretofore, the manufacture of screws with a or other shape of has been impractical, several proposed types requiring casting, which is so expensive as to render manufacturing costs prohibitive. Other types require , but this operation pushes the metal ahead of the broaching tool and so disturbs the distribution of metal as to render the screw head extremely fragile.
IIAs pointed out in said prior application, one of the principal objects of the invention is the provision of a recess in the head of a screw which is particularly adapted for firm engagement with a correspondingly shaped driving tool or screw driver, and in such a way that there will be no tendency of the driver to of the recess when united in operative engagement with each other.
III… since the grooves are of gradually increasing width from their outer edges inwardly toward the axis of the screw, the on the driving tool which are shaped to fit the grooves can be made strong and rigid and such tool will have little or no tendency to cam out of the recess of the screw in driving the same.
Henry F. Phillips and Thomas M. Fitzpatrick, US Patent 2,046,839, Screw, filed 15 January 1935, granted 7 July 1936
The sloping walls are there because a sloping recess can be punched. Casting cost too much and broaching left the head fragile, so the shape had to widen toward the surface in order to let the tool back out, and the cross was the shape that could be driven into a screw head in one stroke without wrecking it. Cam-out came in with that geometry, unintended, and the industry then spent a decade working to reduce it, which is not the way an engineer treats a feature he meant to build. What the slipping did buy, once power drivers came onto assembly lines with unreliable clutches, was a rough ceiling on the torque a bit could deliver before it let go, and the story that this had been the plan all along attached itself afterwards.
The three drives put on the same screws with the same tool. Watch the Phillips bit under load rather than the numbers: the tip lifts visibly in the head before it lets go, and the wear it leaves behind is a rounded tip, which is a shallower wedge, which cams out sooner next time.
Bernard Reiland at Camcar filed the settlement in March 1967. Six lobes, walls standing very nearly true to the axis, almost nothing left in the contact for the driver to climb; it is standardised now as ISO 10664, sold as Torx, and it is in your car and your appliances and, increasingly, your deck screws, for the reason Robertson gave in 1908. The practical residue is short. Cam-out is usually a mismatch, so match the bit number to the screw rather than forcing the nearest one, retire a bit once its tip has rounded, keep hard pressure along the axis while you turn, and where the choice is yours buy square or star.