Morphological glossing presupposes morphological analysis, specifically the identification of morphemes. Rule 8, which actually is a tightening of Rule 3, therefore requires that given a particular L1 morpheme, its gloss will be the same in all contexts.

Rule 8. The morphological gloss represents morphemes, not allomorphs.
Therefore, all the allomorphs of a morpheme share the same gloss; the gloss of a grammatically conditioned allomorph does not contain the feature conditioning it.

For an element (a morph or submorphemic unit) implying value V of category C, two kinds of distribution may be distinguished:

1. The element in question only appears in the context of its conditioner V.
2. The element in question may both cooccur with V and convey V on its own.

The two situations are treated in turn:

1. The straightforward first case may be illustrated with . Latin has different sets of person/number desinences conditioned by the tense/aspect/mood category coded on the verb stem. The gloss of the last morph of the forms in only identifies the values of the categories of person and number (and assumes an unmarked voice). Specifically in #b, it does not say 'PRF.1.SG', although the -i morph does imply the perfect (cf. Lehmann 2014).

   .	a.	lauda-ba-m
   Latin		praise-IMPF(IND)-1.SG(ACT)
   		I praised
   	b.	lauda-v-i
   		praise-PRF(IND)-1.SG(ACT)
   		I have praised

   In , the conditioning factor is a value of an inflectional category speficied in the immediate vicinity of the allomorph and consequently does show in the glossing line. An analogous procedure is possible if the conditioning factor for an inflectional or derivational allomorphy is some grammatical class of the stem. The example of Latin gender is discussed below in #2. Consider here the status suffix allomorphy displayed by the verb in Yucatec Maya. As explained in the section on classes and members, verb stems fall in transitivity classes, and these condition allomorphy in a conjugation category called status, of which Completive is a value. The basic valency class of a verb stem is given with its root; and unless it is changed by some overt process, it is covert, as in .a. It may then be inferred from the status allomorph. It is, nevertheless, not associated with the gloss of the latter, but with the gloss of the stem, as shown in , so the conditioning factor of the allomorphy is recognizable.

   .	a.	hats'-ah-∅
   Yucatec		beat.TR-CMPL-ABS.3.SG
   		beat (past) (it)

   	b.	haats'-nah-ih
   		beat\INTROV-CMPL-ABS.3.SG
   		beat (past)

2. Things are different in .

   .	a.	puer-i	bon-i
   Latin		boy.M-NOM.PL	good-M.NOM.PL
   		good boys

   	b.	puell-ae	bon-ae
   		girl.F-NOM.PL	good-F.NOM.PL
   		good girls

   Apart from motion,² gender is inherent in a noun stem. It is, however, recognizable by the declension suffixes. Nevertheless, the gloss of the morph in question does not contain the conditioning category. The noun forms are glossed as shown, implying that gender is a category of the stem, not of the suffix. Things are different for the adjectives. Gender is not inherent in an adjective stem. The adjectives are therefore glossed as shown. Again, one and the same element receives two different glosses.

   The declension suffix on a noun is an allomorph conditioned by its declension class, including its gender. In some declension classes, it does not even support recognition of the gender. Contrariwise on the adjective, the declension suffix does code gender. The solution adopted here assumes that what appears to be the same element is actually an exponent of two different morphemes. That is, there are two “homonymous” declension suffixes -i in Latin (and likewise for -ae). This is one of the problematic consequences of the 'item-in-arrangement' approach underlying morphological glossing.

The backside of Rule 8 is: Two different forms or formatives only have the same gloss if they are allomorphs. In other words: It is inadmissible to use the same gloss for distinct L1 units, including synonyms. S. the section on mapping.

² ‘Motion’ is the traditional term for gender alternation with the same root, as in Latin lup-us ‘(he-)wolf’ vs. lup-a ‘she-wolf’.