equal, we transfer
the distance _e c_ with the dividers from _j_ and obtain the point _l_.
By setting the dividers on _j_ N and transferring to _l_ N' we draw the
line S' N' on which the locking face of the disengaging pallet will be
located when unlocked. It will be perfectly clear to anyone that through
these means we can correctly represent the pallets in any desired
position.

We will notice that the face Q' M' of the engaging pallet when unlocked
stands at a greater angle to E B than it did when locked, while the
opposite is the case on the disengaging pallet, in which the angle
S' N' F is much less than S N F. This shows that the _deeper_ the
engaging pallet locks, the lighter will the draw be, while the opposite
holds good with the disengaging pallet; also, that the draw increases
during the unlocking of the engaging, and decreases during the unlocking
of the disengaging pallet. These points show that the draw should be
measured with the _fork standing against the bank_; not when the locking
corner of the pallet stands on the primitive circle, as is so often
done. The recoil of the wheel (which determines the draw), is
illustrated by the difference between the locking circle M N and the
face Q M for the engaging, and S N for the disengaging pallet, and along
the _acting_ surface it is alike on each pallet, showing that the draft
angle should be the same on each pallet.

A number of years ago we constructed the escapement model which we
herewith illustrate. All the parts are adjustable; the pallets can be
moved in any direction, the draft angles can be changed at will. Through
this model we can practically demonstrate the points of which we have
spoken. Such a model can be made by workmen after studying these
papers.

[Illustration]

In both the equidistant and circular pallets the locking face S N of the
disengaging pallet deviates more from the locking circle M N than does
the locking face Q M of the engaging pallet, as will be seen in the
diagram. This is because the draft angle is struck from E B which
deviates from the locking circle in such a manner, that if the face of a
pallet were planted on it and _locked deep enough_ to show it, the
wheel would actually _repel_ the pallet, whereas with the disengaging
pallet if it were planted on F B, it would actually produce draw if
locked very deep; this is on account of the natural deviation of the 30deg.
lines from the locking circle. This difference is more