Because a word-internal sequence of two liquids violates the Sonority Sequencing Constraint (see onset: complex onsets), liquids are not expected to be able to form a sequence. However, they seem to do so as /-rl/, in all and only the following words:

Since /r/ precedes the coronal liquid /l/ in the sequence /-rl/ it should have been deleted, but it has not. This means that /r/ and /l/ cannot have been adjacent at the time the historical phonological process of /r/-deletion was operative (see /r/-deletion in simplex words).

Assuming the words ending in -<rl> to have final /-rəl/ in their underlying representation enables one to arrive at an insightful analysis. There are four pieces of independent evidence for this particular underlying representation (which partly equal those for underlying representations ending in /-{l/r}əm/ for words ending in -{l/r}m, see word-final sequences of a liquid and a nasal).

In the first place, the great majority of the forms ending in -rl have a variant ending in -rel, see the table below:

The other way around, not all words ending in -rel have a variant ending in -rl.

Secondly, the five verbs at the end of the table above belong to the second weak class, which receives a principled explanation when they are assumed to end in /-rəl/ (see the Weak II Generalization).

Thirdly, /-l/ may be realized as syllabic, which presupposes the presence of schwa (see distributional evidence that syllabic consonants derive from /ə/ + consonant). Since /l/ in /-rl/ cannot be syllabified in a regular way, the insertion of schwa might be considered a last resort for /l/'s syllabification. This would then be the only instance in Frisian where insertion has this function. Normally, schwa-insertion has the function of facilitating the realisation of a heterorganic tautosyllabic consonant sequence (see schwa insertion (in general)). This renders obligatory schwa-insertion in this particular case highly suspect. Assuming schwa to be part of the underlying representations of the words at hand is, a well-motivated alternative.

Fourthly, pluralization provides us with an indication concerning the underlying representation of the words ending in -rl. The plural ending -s (/-s/) is chosen after words ending in /-əl/, as in leppels/lɛpəl+s/spoons, whereas it is -en (/-ən/) after words in /-l/, for example mielen/miəl+ən/meals. The nouns in Words ending in the sequence -rl all have a plural in -s. This is easily explained by assuming them to end in /-rəl/ in underlying representation (see regular plural formation).

As noted above, for some speakers and/or in some dialects, /r/ never shows up in whatever form of these words, hence it may be assumed to be absent altogether. If waarlcatch, (door, window) fastener, kerlgranule, grain and gurlbelt have the underlying representations /va:l/, /kɛl/, and /ɡøl/, their plurals are expected to be waarlen/va:l+ən/[va:ln], kerlen/kɛl+ən/[kɛln] and gurlen/ɡøl+ən/[ɡøln], respectively, which is borne out by the facts.

With the analysis put forward in this topic then -rl does not have any reality beyond the purely orthographic.