Although there are a large number of rules (Chapter 3) involving consonantal alternations or deletions, it is usually easy to determine the underlying forms of consonants. However, there are some prefixes and suppletive compound initials which must be set up as ending in a final stop or a final nasal, but where this segment is either deleted or assimilated (in point of articulation) in all surface forms so that we cannot directly observe its underlying point of articulation. We thus use //G// for a stop archiphoneme, and //N// for a nasal archiphoneme.
In the case of //G//, we could identify the underlying stop specifically as velar //g//, since this would always produce the correct outputs. However, we play safe and use //G//, since we never actually see the supposed underlying velar (which may or may not be etymologically correct). Similarly, if we substitute //ng// for //N// we will always produce the correct surface forms, but we nonetheless use the more conservative archiphoneme transcription.
Prefixes ending in //G// include derivational noun-class prefixes like MSg/NA //niG-// (see 4.7), Benefactive //-aG-// 10.2 (variant //-wa:G-//), Multiple allomorphs //-w1araG-// (less common than //-w1ara-//) and //-ngaraG-// (alternates with //-ngaran-// 10.4), and various compound initials or minor derivational prefixes (see dictionary entries for //-baG-//, //-burG-//, //-dirG-//, //-lha(G)-//, //-lhagarG-//, //-lhiG-// or //-liG-//, //-malangaG-//, //-malG-//, //-marG-//, //-nguraG-//, //-ragarG-//, //-aga:G-//, //-ambuluG-//, and //-w1urG-//). The //G// always disappears on the surface due to Stop-Deletion P-29), but is needed since following morphemes undergo Hardening P-18 of their initial segments (a rule which requires a preceding stop or nasal), along with other rules likewise requiring such a consonant. Thus //-baG=w1urya-// /-ba=gurya-/ 'to pour over eyes' and //-baG=w2urlha-// /-ba=burlha-/ 'to rub eye', among other compounds, point to //-baG-// rather than *//-ba-// as the form of the compound initial meaning 'eye' (not directly related to the independent noun of this meaning).
Archiphoneme //N// occurs, in the maximally abstract internal analysis of pronominal prefixes (Chapter 9) in the Inverse morpheme //-N-// (requiring ablaut of vowel in some preceding morphemes). It always acquires its surface point of articulation by assimilation to a following consonant except when it is deleted entirely; thus it shows up as surface velar in ANA1Sga //nga-N-w1u-// //nga-N-gu-//(Hardening P-18)/nganggu-/(Nasal Assimilation P-27), but as surface labial in 3Pl/WARA1Sga //nga-N-w2i-// //nga-N-bi-//(P-18) /ngambi-/(P-27). (We generally omit internal hyphens in surface transcriptions of pronominal prefixes.) Other morphemes for which //N// is given in the dictionary as final segment are Gentilic //nuN-// (used mainly with clan names 14.11) and these minor compound initials or derivational elements: //-lhaN-//, //-lhangaN-//, //-laN-//, //-ngaN-//, //-raN-//, //-ambaN-//, //-albunguN-//, //-aN-//, and //-ilbuN-//. In all of these we have a nasal segment which acquires surface point of articulation from the following consonant, or else is deleted by the usual rule which deletes (certain) nasals in the relevant combinations (Nasal-Deletion P-30). For the majority of prefixes, compound initials, and the like we do not need consonantal archiphonemes: either the morpheme clearly ends in a vowel, or it ends in a consonant which can be uniquely specified. Many compound initials, for example, are merely special uses of stems which also occur as independent nouns, and in the latter function it is easy to determine the final segments.