a-stems divide into masculines and neuters. You may think of hestr “horse” as the standard paradigm for strong masculine nouns. Characteristic of the a-stems (including its subvarieties the ja- and wa-stems) is that masculines have nominative plurals in -ar, while both masculines and neuters generally have a genitive singular in -s. Remember, the -a- in the nominative plural of masculines of this class is in fact the thematic vowel.

The -r at the end of the word's lexical form is the nominative singular suffix, corresponding to -us in Latin dominus. It disappears in the accusative, so if we were to construct the shocking headline “horse eats horse” it would read “hestr etr hest”. The genitive ending is -s, just as it is in Modern English, though there we spell it with a preceding apostrophe: horse's. The dative singular for this class of noun ends in -i. It may be noted that these singular forms are like the Old English forms of the same class except for the nominative singular suffix -r, the loss of the e in the genitive (compare stānes) and a very slight difference in the dative form, which in Old English is -e.

In the plural, the nominative and accusative endings differ from the Old English, though not terribly much. The -ar of the nominative is actually just the North Germanic reflex of the Proto-Germanic az-ending that turned into -as in West Germanic. If you think about it, /z/ and /r/ are very similar sounds (both are alveolar consonants, but /z/ is a fricative and /r/ a trill). The -r has gone missing from the accusative plural, however, making it identical to the genitive plural. In Old Norse as in Old English, there are two certainties in nominal life: genitive plural ends in -a, dative plural in -(u)m. Old Norse knows exceptions to this rule (cf. kné below), but it still pays off to pretend.

The paradigm for hǫfundr “chieftain” is identical to that of hestr except in one particular: it substitutes -ar for the genitive singular ending -s. A range of common and proper nouns will do this, and many can take either form.

Harmr “sorrow” does exactly the same thing as hestr as far as stems are concerned, but it illustrates one additional sound law known as labial mutation. This law operates in all parts of speech and states that an a preceding a syllable that has a u or w in it is rounded to ǫ. This is because both /u/ and /w/ are pronounced with liprounding (hence the term “labial mutation”). The letter w, however, is not normally used in the Old Norse or any other Scandinavian alphabet, where we find v instead to represent the labial approximant. In the dative plural, the um-ending triggers anticipatory liprounding, turning a into ǫ, *harmum into hǫrmum. The rule also applies to a few other sounds, such as á and e, but a > ǫ is the one you will actually encounter. Unfortunately, labial mutation also takes place in words in which the u or w that caused it has since been lost, so that students may want to memorize where in the paradigms the phenomenon is to be expected. Conversely, it is found throughout hǫfundr because u is there a reliable part of the root. The harmr is more helpful in demonstrating where in the paradigm liprounding is triggered by elements outside the root.

Himinn “heaven” displays two further variations on the same declension pattern. The first is in the nominative singular ending. Originally, this would have been -r as in the other masculine paradigms of this class: *himinr. However, as is elsewhere demonstrated by the development from *mannr to maðr, early speakers of North Germanic found the nr-cluster difficult to pronounce, and so the -r of *himinr was assimilated to the preceding n, yielding himinn. In the same way, the cluster -lr became -ll (ketill) and -sr became -ss (íss).

The other variation is a process called syncope, or loss of a medial vowel. Dissyllables (two-syllable word forms, also called disyllables or bisyllables) resist adding a third syllable when possible, so the second vowel will typically drop out if an inflectional ending containing a third vowel is added. This was possible because stress was primary, so the second syllable was already pronounced weak; adding a further syllable was just the excuse needed to ignore the vowel—and thus the syllable—altogether: *himini > himni. With the exception of these mutations, however, himinn still inflects in exactly the same way as hestr. No vowel is lost if the word's first syllable is long or if the vowel in question is followed by two consonants (e.g. in the adjectival dative singular feminine form gamalli).

Barn “child” is our first neuter, and it introduces three further complications compared to the base-pattern. The first is an Indo-European given, namely that neuter nominative and accusative forms are always identical within the same number: thus for both cases we have barn in the singular, bǫrn in the plural.

But those nominative and accusative plural forms strike the eye for two additional reasons. Firstly, neither has an ending. This should come as no surprise to students of Old English, who will remember that some a-stem neuters in that language have no plural endings for these cases either (namely long stems like word and some dissyllables), while others (like scip) end in -u. In Old Norse, all neuter a-stems, which is to say all strong neuters, have endingless nominative and accusative plural forms.

As in Old English, however, -u is the historically antecedent nominative and accusative plural ending for neuter a-stems. (We will see below that this ending still occurs in weak neuters.) Accordingly, in barn we have our first example of labial mutation (specifically u-mutation) occurring where the sound that gave rise to it no longer survives.

If you have taken Old English, you may recall a great deal of fuss over the concept of stem length. It is time now to revisit it ever so briefly. The four paradigms in this section demonstrate two variants of the same ja-declension for each of two genders (remember, there are no feminines among the a-stems). While all four follow their gender-respective version of the basic a-stem pattern, the short stems (those whose root syllables have a short monophthong vowel followed by no more than one consonant, or a long vowel or diphthong followed by no consonants at all, i.e. niðr and kyn among the paradigms here provided) have a -j- intervening between the root and ending wherever a back vowel (a or u) follows, i.e. in all plural forms for masculine nouns, and just the genitive and dative for neuters. Long stems of both genders (i.e. stems with a short vowel plus two or more consonants or a long vowel or diphthong followed by a consonant) have an -i- across the cases of the singular (including the genitive). Some grammars here speak of a separate subclass “īa-stems”; we will refer to them simply as long ja-stems. The disappearance of -j- from the genitive and dative plural of long-stemmed hirðir “shepherd” but not ríki “power” is due to the fact that squeezed in between a long syllable and a back vowel or ę (representing short /æ/, a phoneme that merged with e in the early literary period), medial -j- disappeared (as in hirða, hirðum) unless immediately preceded by g or k (as in ríkja, ríkjum).

Almost no other word declines like niðr “son, kinsman”, as most masculines of this sort had assimilated to the i-declension by the time they were written down. Like ríki are a fair number of neuter nouns in -i. Because -i is also the dative singular ending, this paradigm comes closest of any strong group of nouns to being indeclinable altogether. Cf. weak type (c) below, whose members also end in -i but have fully indeclinable singular forms and are exclusively feminine.

wa-stems are simply a-stems that originally ended in -w. In Old Norse, the -w is represented by v, but it disappeared except where it both is followed by a or i and itself follows a short stem (sær), g (sǫngr, hǫgg), or k. In some places, the -v- was reintroduced by analogy; hence dative plural sævum. Where the original -w immediately followed a vowel (and hence where the surviving v follows a vowel), the genitive singular ending of the word is usually -ar rather than -s; however, there is some contamination between the forms, so that sæs is also attested.

Kné “knee” is really an intruder among the wa-stems, as it derives from an Indo-European u-stem (cf. the Latin cognate genu). It was made to conform to the pattern of inflection of the wa-stems in the Germanic languages, but by the time it was written down in Old Norse only the genitive singular ending suggests a connection with the a-stems at all, let alone the wa-variant (but cf. Old English cnēow). The word for tree, whose Norse reflex is tré (Old English trēow), underwent the same process, and there are one or two words that behave in the same way, but really the pattern is very rare. The student had best shrug and move on.

The ō-stems correspond to the PIE ā-stems seen in the Latin class one nouns (vita), and they are all feminines. In this they contrast with the corresponding class in PIE, which had some masculines (cf. agricola), and they form a perfect complement to the Germanic a-class, which has no feminines. This class is characterized by a zero (i.e. endlingless) nominative singular; a genitive singular as well as a nominative and accusative plural in -ar; and u-mutation in the nominative singular in stems whose vowel is eligible for labial mutation.

That instance of u-mutation occurs because in an earlier stage of the language these nouns had a nominative singular in -u (like Old English giefu): *þarfu > *þǫrfu > þǫrf. The accusative singular vowel was rounded by analogy with the nominative, and the rounded vowel of the dative derives from a one-time u-ending there as well. Only in the dative plural is the cause of the mutation still visible. The phenomenon never spread to the other forms, where an a always follows the root, discouraging liprounding in adjacent vowels if no v (from /w/) intervenes (cf. genitive singular dǫggvar). the The paradigm thus contains an equal number of mutated and nonmutated forms, in which the genitive singular and the dative plural stand out in their respective number. It is also worth observing that the genitive singular, nominative plural, and accusative plural forms of this class are identical in form. The difference between the two paradigms is the dative singular in -u, which occurs in many dissyllables in -ing or -ung as well as a variety of other nouns of this class, including hlíð “slope”, laug “bath”, and mjǫll “freshly-fallen snow”. Of these, all except the words in -ing or -ung could also be conjugated like þǫrf.

Á “river” shows no evidence of u-mutation. Although the sound represented by á was subject to labial mutation into , the two sounds merged in the early literary period, after which they both surfaced with the spelling á.

As with the ja-stems, so also in the jō-stems -j- remains in full only following short stems, while in long stems it survives as -i- wherever the expected ō-behaviour is to have no ending, i.e. in the accusative and dative singular. Take note of the variation in the dative singular of short stems (-ju versus no ending). Most short stems decline like ben, but some, including egg “edge” and hel “hell” decline like ey, as do many personal names (indeed, hel itself started out as a personal name).

Other than the paradigm dǫgg “dew”, the only common words of the wō-stems are þrǫng “pressure” and ǫr “arrow”; the poetic term bǫð “battle” also belongs here. The optional -u of the dative singular is comparatively rare.

i-stems divide into masculines and feminines, which largely share the same endings except in the accusative plural, which for masculines ends in -i, while feminines end in -ir. In effect, verbs of this class behave like a- and ō-stems but show a different thematic vowel in the nominative and accusative plural, and normally with a zero dative singular ending even in masculines. However, note the following:

• There is variation in the genitive singular ending of masculine nouns between -s (gests) and -ar (staðar);
• there is some contamination from the a-stem ending -i in the dative singular of masculines, yielding forms like gesti, while feminines occasionally take a -u in the dative singular (rǫstu).

Other than that, patterns are largely as we have by now come to expect, including u-umlaut where applicable, namely always in the dative plural where the stem vowel is a, and with feminines also in the nominative, accusative, and dative singular.

Unlike in Old English, the u-stems in Old Norse are all masculine. They exhibit three notable sound changes:

• Breaking, triggered when a Proto-Norse (PN) stem with e as its vowel is followed by a syllable with either a (a-breaking or u (u-breaking). Before a, e broke into ea by anticipation, which in turn became ja, as in the genitive forms skjaldar and skjalda. Where the combination was e+u in an earlier stage of the language, the e appears as jǫ in literary times: *skeldur > *skjǫldur > skjǫldr. Neither type of breaking occurs if the stem vowel is preceded by l, r, or v.
• u-mutation, also seen in other classes and described for the a-stems above, compounds breaking in the u-stems. In fact, breaking before u seems initially to have resulted in ja just as it did before a, but later, when u-mutation occurred, that ja turned into jǫ. Since both phenomena are triggered by a following u, the resulting jǫ incorporates both the breaking and the liprounding. Where breaking did not occur (namely because the original root vowel was a not e, as in vǫllr “field” or fǫgnuðr “joy”), the vowel would still be ǫ courtesy of u-mutation.
• A third sound change here visible is i-mutation. This is the same phenomenon that causes the vowel variation between the Modern English forms goose and geese, foot and feet. It specifies that a back vowel followed by an i or j in the next syllable will be fronted, i.e. turn into a vowel with the same qualities as before except pronounced with the front of the tongue raised where the original vowel was pronounced with an elevated back of the tongue. For more about this, see my write-up on weak verbs. Important to remember here is that it turns the one-time a of the dative singular and nominative plural of many u-nouns into e.

Fǫgnuðr represents common and proper nouns in -uðr (-aðr): these are not subject to i-mutation, so the dative singular and nominative plural retain their back vowels despite the following -i.

Also known as athematic stems (or root-stems), these masculine and feminine nouns have no thematic vowel. Endings are largely as in the a, ō, or u-stems, but the nominative and accusative plural ending is -r. Where stems end in -l or -n, however, this -r has disappeared by assimilation. i-mutation occurs in the nominative plural (from PG *-iz) and the accusative plural (by analogy with the nominative plural). Originally, it also occurred in the dative singular of masculines (as a consequence of the -i ending) and the genitive singular of feminines, but these forms had largely given way to nonmutated forms by the literary period, with the dative form fœti a rare example of surviving i-mutation while the genitive bœkr is most commonly found as bókar instead. This class is the origin of present-day English gradation-plurals (foot : feet; man : men). It is not particularly large, but its members occur very frequently.

Relational nouns like father and mother are of course commonly used, which means that their origins can be traced quite closely, but also that they tend towards irregularity. The oblique cases of the singular are particularly variable: faður, fǫður, and fǫðr are all common spellings of the accusative, genitive, and dative forms of faðir “father”, especially later in the period of written Old Norse; in the same way, each of the oblique cases of bróðir can be either brœðr or bróður, despite the historically justified simplification in the paradigm here presented. Although this Indo-European class ultimately derives from agent nouns, only family nouns are found in the Old Norse class. This class comprises faðir, móðir, bróðir, systir, and dóttir, but not sonr, which is a u-stem.

This class is populated by masculine present-participle stems: -nd- is the marker of the present participle in the older or more conservative Germanic languages (even Modern English friend ultimately derives from the present participle of Old English frēogan “to love”: a friend is one who loves, a frēond). Accordingly, you may want to think of gefandi as “giving person, giver” and of bóndi (or búandi) as “farming person” or “dwelling one”.

Now that you've got the strong nouns under your belt, you will find that there is not much to the weak nouns. There is only one top-level weak noun declension, known as the n-stems, even if it does subdivide into three subclasses. You will find that a great number of masculine and feminine nouns follow the basic an- and ōn-declensions, and none of the various mutations should present any real difficulty.

an-stems are mostly masculines and some neuters. Bogi “bow” demonstrates the default weak masculine declension. Since its plurals are like those of masculine a-stems, it is the singular oblique cases that set this type apart from other declensions.

Gumi and bryti demonstrate two minor variations. Some plurals show an -n- in their ending, which represents an older form of the genitive plural which has been lost in most words (like bogi). Where it survived, however, it could spread to other cases (as in gumi). The second variation (known also as the jan-stems) has a -j- in all forms except the nominative singular. The latter phenomenon is limited to stems ending in certain morphemes, of which -ingi, -viri, and -virki are the most common.

This small class consists of feminines in -i only (compare the strong neuter long ja-stems as represented by the ríki paradigm above). Since most of its members lack a plural and are indeclinable in the singular, you will simply want to know not to be surprised to encounter the occasional feminine noun without the typical feminine bits. Where a plural does exist it follows ō-stem declension.