ow of the ridges on either side. No angle is present as
the ending ridge does not abut upon the curving ridge which envelopes
it.

[Illustration: 335]

[Illustration: 336]

[Illustration: 337]

[Illustration: 338]

[Illustration: 339]

Figure 337 is a plain arch because the dot cannot be considered a
delta as it is not as thick and heavy as the surrounding ridges.

Figure 338 is a tented arch consisting of two ending ridges and a
delta. The short ending ridge is considered a ridge because it is
slightly elongated and not a mere dot.

In figure 339, the only question involved is where to stop tracing.
The rule is: _when tracing on a ridge with an upward trend, stop at
the point on the upward trend which is nearest to the right delta_. X
is the point in this pattern.

In figure 340, the question involved is also one of tracing. In this
pattern, the tracing is not on a ridge with an upward trend. The
tracing, therefore, is continued until a point nearest to the right
delta, or the right delta itself, is reached. This tracing is a
meeting tracing.

[Illustration: 340]

There are a few constantly recurring patterns which, though not
questionable or doubtful as they appear, present a peculiarly
difficult problem in classifying. The patterns referred to are usually
double loops, though accidental whorls and loops sometimes present the
same problems. The difficulty arises when a loop is so elongated that
the recurve does not appear until near the edge of a fully rolled
impression or an impression that is rolled unusually far, as in
figures 341 to 344.

[Illustration: 341]

[Illustration: 342]

[Illustration: 343]

[Illustration: 344]

Figure 341, if classified as it appears, would be an accidental whorl.
Figures 342 and 343 would be double loops, and illustration 344, a
loop. It will be observed that these prints are rolled more fully than
normal. If, however, the next time the prints are taken, they are not
rolled quite so far, the patterns would require a very different
classification, and would show no indication of any need for
referencing to their true classification. The result would be a
failure to establish an identification with the original prints. The
only way in which such an error may be avoided is to classify such
impressions as they would appear if not so fully rolled, and to
conduct a reference search in the classification which would be given
to the prints when rolled to the fullest ext