Senatus libertatem his civitatibus dedit. (Liv. Hist. 33. 34)
‘The Senate gave freedom to these states.’

In addition to the nominative, accusative and dative, which were the cases assigned to the verb’s principal arguments, i.e. the phrases designating the main participants in the proposition expressed, the Latin morphology included a case which typically related one noun to another, for example in a possessive relationship, and a case which marked the noun phrase as having an adverbial role in the clause. The former is referred to as the genitive case and the latter as the ablative. Examples of both are given below:

Genitive case:

cervi cornua
‘the stag’s antlers’

odium servitutis
‘hatred of slavery’

spem victoriae
‘hope of victory’

Ablative case:

gladio manum praecidit (Cic. Inv. 2. 59)
‘(he) cut off his hand with a sword’

Syracusis illa nocte profugit (Cic. Verr. 2. 2. 17)
‘(he) flees Syracuse during the night’

For nouns that occurred as the complement of a preposition, Latin normally used either the accusative or the ablative case, depending on the preposition or its specific meaning. For example, ad ‘to(wards)’, inter ‘among/during’ and per ‘for/per’ selected the accusative, while cum ‘with’, ex ‘from’, prō ‘for’ and dē ‘of/about’ selected the ablative. The prepositions in ‘on/at’ and super ‘on/above/beyond’ selected either case depending on the sense in which they were being used. Examples involving various prepositions are given below:

Preposition + accusative case:

Ad hostem contenderunt.
‘They marched against the enemy.’

Ligarius in Africam venit.
‘Ligarius entered Africa.’

inter cenam
‘during dinner’

Preposition + ablative case:

Cum amicis deliberavi.
‘I consulted with friends.’

Vivo in urbe.
‘I live in the city.’

homo de plebe
‘man of the common people’

Latin also had a vocative case, which was used for direct address. However, this was marked only in a minority of nouns and never in the plural. Moreover, by its nature, it did not express any syntactic relation and its existence can be regarded as being peripheral to the question of how the Latin nominal system evolved. In Spanish, it survives only in a few fossilized proper nouns, such as Santiago (< Old Sp. Santi Yagüe < Latin Sāncte Jācōbe, voc. sing. of Sānctus Jācōbus).

Latin nouns can be grouped into into classes or declensions, according to the pattern of their case endings. In the Classical variety of the language, there were five such classes, but for later spoken Latin it is more appropriate to think of three just declensions, as the relatively small fourth and fifth declensions were absorbed into the other three. These three main declensions are illustrated in Table 1 below, using the examples of tĕrra ‘land’ and lŭpus ‘wolf’ in the first and second declensions and of pater ‘father’ and pānis ‘bread’ in the third. The latter declension was in fact characterized by the unpredictable nominative singular forms of many of its nouns, which complicate to an extent the evolution of the system overall, as is discussed in more detail in Section 1.5. The table excludes the vocative case, in line with the remarks in the previous paragraph.

As can be seen from Table 2 below, the equivalent noun paradigms of Spanish are much simpler:

In camparison to its Latin ancestor, the modern noun system reflects two key changes. In the first place, while the distinction between singular and plural has survived – and in fact become more clearly marked – morphological case has been entirely abandoned. Secondly, each class has developed a unique stem vowel, i.e. the vowel that comes immediately after the lexical root, viz. /a/ in Class 1, /o/ in Class 2 and /e/ in Class 3. The vowel /a/ was already reasonably prominent in the endings of the Latin first declension, but the stem vocalism in the second and third declensions was, to begin with at least, fairly inconsistent. As regards the /e/ that triumphed in the third declension, it can be noticed that this has been lost in the singular of nouns like pan but retained in both the singular and plural of nouns like padre. This dual outcome reflects the operation of the (relatively late) sound change apocope of final /e/, which only affected final /e/ if it followed an ungrouped dental or alveolar consonant.

The evolutionary pathway linking the old Latin system to the modern Spanish one is discussed in the following sections.

1.2 Increased use of prepositions
The Classical Latin system illustrated in Table 1 already exhibits a fair degree of syncretism, in the sense that in each of the columns one or more forms appears in two or more cells. With the passage of time, this syncretism only increased, owing to the loss of final /m/, the distinguishing mark of the accusative case in the singular, and the vowel mergers occasioned by the disappearance of length distinctions. In the first declension, for example, the singular forms of the nominative, accusative and ablative came to have an identical pronunciation, e.g. [ˈtɛɾɾa] for tĕrră, tĕrrăm and tĕrrā, while in the second declension the ending of the nominative singular became identical to that of the accusative plural, both converging in /-os/, and in the third declension the nominative singular of regular nouns like pānĭs, which had always shared their form with the genitive singular, came to have the same phonetic realization as both the nominative and accusative plurals, the ending in all of these forms now being /-es/.

An important consequence of the syncretism that arose among the singular forms was that the ablative case became indistinguishable from the accusative case (in the singular, but not in the plural). This was brought about by the fact that, in addition to the above-mentioned convergence of items like tĕrră and tĕrrā in the first declension, the -ō and -ŭm endings in the second declension merged into /-o/ while -ĕ and -ĕm in the third declension merged into /-e/. Once these mergers had taken place, the ablative and accusative singulars were marked in exactly the same way in each of the three declensions, i.e. by the presence of the bare stem vowel /-a/, /-o/ or /-e/. From that time onwards, only the tiny minority of speakers who were schooled in the precepts of Latin grammar would have retained any sense of an ablative-versus-accusative case distinction in the singular.

In a parallel development, spoken Latin gradually came to rely on prepositions to express grammatical functions other than the core syntactic roles of subject and direct object. In reality this process represented no more than the generalization of a pattern that had always been available, in the form of the ‘preposition + ablative’ and ‘preposition + accusative’ constructions mentioned in 1.1. In the singular category, these two constructions were (or came to be) a single structure, given the above-mentioned merger of the ablative and the accusative. Over time, this merged structure must have extended its domain of use beyond the functions originally associated with the prepositional ablative and accusative and into the semantic territory traditionally reserved for the genitive and dative cases, eventually outcompeting these latter cases and causing them to fall into obsolescence. For example, to express the meaning ‘of this woman’ it can be surmised that the replacement shown below must have occurred in speech:

Classical form:
ĭstīus mŭliĕris (gen. sing. of ĭstă mŭliĕr)

Replaced in speech by:
[de ˈesta moˈljɛɾe] = dē ĭstam mŭliĕrem (acc.) or dē ĭstā mŭliĕre (abl.)

Examples from early Romance suggest that the merged ablative/accusative could occasionally be used without a preposition to express the genitive and dative functions (see Salvi 2011:320), but the overwhelming trend was towards the use of prepositions for all case-related functions other than subject and direct object.

The logical outcome of this process would be the replacement – in the singular – of the system illustrated in Table 1 by a binary system, with the following components:

(i) a continuation of the nominative which, as the subject case, was not susceptible to being replaced by a construction involving a preposition;

(ii) a continuation of the now merged ablative and accusative, characterized morphologically by the absence of any specific marker beyond the bare stem vowel associated with the relevant declension, i.e. /-a/ in the first declension, /-o/ in the second and /-e/ in the third.

The merged case referenced in (ii) is conventionally designated using the label oblique, a practice that is followed here.

Turning now to the plural category, it seems clear that a merger here between the accusative and the ablative was not possible, given the phonetic disparity between the relevant endings: -ās, -ōs, -ēs (acc.) versus -īs, -īs, -ĭbus (abl.). On the other hand, of these two sets of endings only the accusative ones survive into Spanish. It thus seems likely that the binary system envisaged above for the singular category was extended analogically into the plural, with speakers identifying the accusative (rather than the ablative) as the equivalent case in the plural to the merged accusative/ablative of the singular. According to this line of reasoning, spoken Latin would have replaced phrases such as ĭstārum mŭliĕrum (gen. plu. of ĭsta mŭliĕr) with dē ĭstās mŭliĕres (dē + acc. plu.), by analogy with the corresponding construction in the singular.

An additional point to note in respect of the plural is that, according to Aebischer (1971) and other authors, the nominative ending in the first declension, viz. -ae (= /-e/), was replaced by /-as/ at an early date in many varieties of Latin, including the Latin of the Iberian Peninsula. Menéndez Pidal (1987: 208) cites a second-century Iberian inscription that illustrates precisely this phenomenon, filias appearing in place of the Classical Latin nominative plural form filiae:

filias matri piissime posuerunt
‘the daughters put up [this memorial] to a most pious mother’

The replacement of plural -ae by /-as/ will be significant for understanding how the binary system developed into the modern nominal system which lacks overt case entirely (a residue of case remains of course in the pronominal system).

1.3 Binary case system of late spoken Latin
Taken together, the foregoing developments can be hypothesized to have resulted in the system illustrated in Table 3 below, which uses phonetic representations rather than orthography, given that by the period in question the relationship between the spoken language and the written one had become opaque.

Note that, in relation to Spanish, the system shown in the above table has to remain within the realm of plausible hypothesis, because even the earliest attestations of Old Spanish show no residue at all of the Latin nominal case distinctions. The hypothesis that such a system existed within the history of Spanish is however supported by Old French, which did retain a constrast between nominative and oblique cases, among nouns from both the second and the third Latin declensions:

Second declension:
veisins ‘neighbour’ < vicīnus (nom. sing.)
veisin ‘neighbour’ < vicīnum/ō (acc./abl. sing.)
veisin ‘neighbours’ < vicīnī (nom. plu.)
veisins ‘neighbours’ < vicīnōs (acc. plu.)

Third declension:
chiens ‘dog’ < canis (nom. sing.)
chien ‘dog’ < canem/e (acc./abl. sing.)
chiens ‘dogs’ < canēs (nom./acc. plu.)

1.4 Reanalysis of /-s/ as a plural marker
Assuming that a system such as the one shown in Table 3 did indeed emerge in the history of Spanish, a clue as to how the modern system developed from it is provided by the prevalence of the suffix /-s/ in the plural category, where it appeared in five of the six possible forms. A plausible hypothesis is that this suffix, which in Classical Latin had not been an exponent of number, was reanalysed as a plural marker. Under this scenario, speakers came to treat nominal forms in /-s/ as plurals and, as a consequence, deleted the final /s/ from those singular forms, all nominative, in which it occurred. In this way, items such as [ˈlopos] (< lŭpus ‘wolf’) and [ˈpanes] (< pānĭs ‘bread’) would have lost their final /s/ and hence come to be pronounced identically to their oblique counterparts.

Either as a trigger for the foregoing process or as a by-product of it, analogical pressure must have caused the increasingly anomalous second-declension nominative plural ending /-i/ to be replaced by /-os/, thereby bringing the plural subparadigm in that noun class into line with what existed in the first and third declensions. This development, combined with the process outlined in the previous paragraph, led to complete merger of the nominative with the oblique case in the first and second declensions, as well as near merger in the third declension.

In the latter class, although the nominative and oblique were indistinguishable in the plural, the ending for both cases being /-es/, quite a few nouns had a singular paradigm in which the nominative was not formally identical to the oblique form plus /-s/. For example, the attested nominative singular of the word for ‘man’ is hŏmō, whereas the oblique singular form must have been [ˈɔmene], from acc./abl. sing. hŏmĭne(m), the source of Old Spanish omne/omre and, ultimately, modern hombre. Among nouns of this latter type, a formal distinction between the nominative and oblique cases would be expected to survive the reanalysis of /-s/ as a plural marker, rather than disappear as a consequence of it. However, if a case distinction did persist in this subgroup of third-declension nouns, it did not continue into Old Spanish, implying that even these irregular nominative singular forms were eventually lost or merged with their oblique counterparts.

Merger can presumably be posited for the specific nominative singular form [ˈpateɾ] shown in the table above, as this would become identical with its oblique counterpart if it is assumed that metathesis (i.e. transposition) of [e] and [ɾ] occurred in speech. However, a simple account like that is not available for most of the irregular third-declension nominative singular forms, which very often were shorter than the other singular forms in their paradigms. Thus in cases like that of hŏmō, mentioned above, there is no straightforward route to merger with the corresponding oblique form. For nouns like this, simple loss of the nominative singular form in late spoken Latin or early Ibero-Romance seems as good an explanation as any, but see 1.5 immediately below.

1.5 Possible expansion of the short nominative forms
Another possibility (see Penny 2002: 118) is that short nominatives like hŏmō suffered the same fate as glīs ‘doormouse’ and grūs ‘crane’. These two items, both short nominative singulars, are mentioned in the Appendix Probi, which condemns their replacement by the longer, analogical forms gliris and gruis. The latter result from the application of the common third-declension nominative singular suffix -ĭs (= /-es/) to the stem found in the accusative, genitive, dative and ablative forms of the relevant nouns, viz. glīrem, glīris, grŭem, grŭis etc. If hŏmō was similarly expanded, specifically to [ˈɔmenes], modelled on hŏmĭnem, hŏmĭnis etc., merger with the corresponding oblique form would have occurred naturally when /-s/ was reanalysed as a plural marker and eliminated from the nominative singular. The same explanation would also be applicable to items like fōns ‘spring’, leō ‘lion’ and rēx ‘king’, which were each one syllable shorter than their oblique form, viz. [ˈɸwente] < fŏnte(m), [leˈone] < leōne(m) and [ˈɾeɟe] < rēge(m). However, in the majority of instances, including that of hŏmō, the putative analogical nominative singular form is unattested in the written record, meaning that any account along the proposed lines, though eminently plausible, remains confined to the realm of speculation.

1.6 Summary
Aside from the uncertainty detailed in the previous section, the basic narrative outlined in 1.2 to 1.4 probably represents a good approximation as to how a nominal system was formed in which there were no case distinctions at all. To recapitulate, the multi-case system of the literary dialect reduced over time to a binary system in which only the ‘structural’ cases identified in theoretical syntax, viz. nominative and accusative, were expressed in the morphology. This diminished formal system was precarious, however, given that the case opposition was expressed primarily by the presence or absence of the suffix /-s/, which also happened to be the principal expression of plurality. Perhaps inevitably, /-s/ subsequently became specialized in just one of these functions, specifically that of plural marker, its role as an exponent of nominative case disappearing as a consequence.

In terms of what survives in the modern language of the Latin morphology, the following can be identified:

(i) the three stem vowels /a, o, e/, the third of these being apocopated in singular forms in which it followed an ungrouped dental or alveolar;

(ii) the suffix /-s/, now reanalysed as a plural marker.

The remainder of system has fallen victim to a combination of sound change, analogical adjustment and a general typological shift from a ‘synthetic’ case-based language to an ‘analytic’ one based on the use of prepositions to express grammatical functions.

2. Consolidation of gender marking
Most nouns in the Vulgar Latin second declension, the /-o/ class, were masculine and virtually all of those in the first declension or /-a/ class were feminine. This correlation between grammatical gender and phonology was progressively deepened and by the Old Spanish period barely any nouns in -o or -a deviated from it.

This result was brought about by the operation of two processes. Firstly, some nouns that ended in /-o/ but were in fact feminine underwent a change in their final vowel. For example, the feminine noun sŏcrum ‘mother-in-law’ appears in Spanish as suegra, the etymological final /-o/ having been replaced by the analogical ending /-a/. This particular case is actually referenced in the Appendix Probi, which commands socrus non socra.

In contrast to what happened in the sŏcrum/suegra type of case, some feminine nouns in /-o/ were reassigned to the masculine gender. This was particularly common with names for trees. Latin ŭlmum, for example, a feminine noun meaning ‘elm’, comes into Spanish as the masculine olmo.

While the relationship between form and gender was, or became, almost completely transparent among nouns in the /-o/ and /-a/ classes, the relationship was fundamentally opaque among the nouns of the Vulgar Latin third declension. In principle, these nouns had singular forms in /-e/ although, following the generalized apocope of final /e/ after ungrouped dentals and alveolars at the end of the Early Middle Ages, many of them came to end in consonant. A subset of both types, i.e. nouns in /-e/ and those that ended in a consonant, underwent gender hypercharacterization, a process whereby their hitherto gender non-specific form was reshaped in accordance with their gender. Two examples are given below:

pŭppem (fem.) > popa ‘stern (of ship)’
cŏchleāre (neuter) > Old Sp. cuchar (fem.) > cuchara ‘spoon’

3. Demise of the Latin neuter
If the loss of case represents the pre-eminent diachronic event in the history of the Spanish noun system, the disappearance of the Latin neuter gender is perhaps the second most important change. In a sense, the Latin neuter was (or became) structurally superfluous, given that the class of nouns that bore the neuter gender was not semantically distinctive. In particular, the large category of inanimate count nouns was split, fairly randomly, between neuters, masculines and feminines. For example, the words for ‘arm’, ‘oak’ and ‘spoon’ were neuters (viz. bracchĭum, rōbŭr and cŏchleāre) but the semantically comparable nouns ‘foot’, ‘ash tree’ and ‘knife’ were masculine, feminine and masculine respectively (viz. pēs, fraxĭnus and cŭlter). Latin neuters did, however, exhibit two very important morphological properties. Firstly, their nominative and accusative forms were always identical. Secondly, their nominative/accusative plural always ended in -a, e.g. fŏlia ‘leaves’, ŏpĕra ‘works’ and cŏrpŏra ‘bodies’.

Neuters in the second declension had nom./acc. singular forms in -um (= /-o/) and were thus naturally reanalysed as masculine nouns, acquiring analogical plurals in /-os/. For example, modern castillo ‘castle’ (masc.) descends from neuter castĕllum, the plural of which was originally castĕlla, a form which must at some point have been replaced by †castĕllos. A number of other familiar Spanish words belong to the same category:

vīnum ‘wine’, vīna > vino, †vinos
prātum ‘meadow’, prāta > prado, †prados
gaudium ‘joy’, gaudia > gozo, †gozos
fīlum ‘thread’, fīla > hilo, †hilos

An analogous diachronic trajectory was followed by the much smaller class of neuters that belonged to the fourth declension, a notable example being cŏrnū ‘horn’, which gives modern cuerno. This type of noun originally had a final /u/ in its singular nominative/accuative form, but this underwent regular sound change to /-o/, paving the way for merger with the neuters of the second declension. In the specific instance of cŏrnū, this merger was foreshadowed in Classical Latin, in which alternative, second declension-style forms of this noun already existed. In particular, besides cŏrnū, the form cŏrnum is also attested as the nominative/accusative singular.

The second declension also provided a home for third-declension neuters with singular nom./acc. forms in -us (= /-os/), such as pĕctus ‘chest, breast’ and tĕmpus ‘time’. The final /s/ of these singular forms must have been lost as part of the reanalysis of /-s/ as a plural marker and, like the neuters mentioned above, the plural forms of these nouns were refashioned with the regular masculine ending /-os/:

pĕctus, pĕctŏra > pecho, †pechos
tĕmpus, tĕmpŏra > tiempo, †tiempos

On the other hand, some neuters that could have become second-declension masculine nouns in fact evolved through their plural in -a, and hence were reanalysed as feminine nouns of the first declension. The following are some common examples:

fŏlia (pl. of fŏlium) > hoja ‘leaf’
lĭgna (pl. of lĭgnum) > leña ‘firewood’
pĭgn(ŏ)ra (pl. of pĭgnus) > Old Sp. peyndra/pendra > prenda ‘pledge, garment’
vōta (pl. of vōtum) > boda ‘wedding’
mōra (pl. of mōrum) > mora ‘blackberry’

Third-declension neuters with nom./acc. singular forms in /-e/, such as rete ‘net’ and mare ‘sea’, looked like any other third-declension noun once they had acquired analogical plurals in /-es/. However, they lacked any overt phonological cue as regards gender and hence were reanalysed somewhat arbitrarily in this regard, a fact reflected in the cross-linguistic variation in terms of the gender of such nouns in modern Romance (compare e.g. Spanish el mar with French la mer).

A significant subset of the third-declension neuters originally had short nom./acc. singular forms, which were expanded analogically in spoken Latin. In most cases this process involved the acquisition of a regular third-declension ending in /-e/, although any resultant forms that were greater than two syllables in length subsequently underwent intertonic vowel loss and consequent phonetic adjustment:

lacte (attested for lac) > leche ‘milk’
[ˈnomene] (for nōmen) > [ˈnomne] > [ˈnombɾe] nombre‘name’
[ˈiŋgwene] (for inguen) > [ˈiŋgne] > [ˈiŋgle] ingle ‘groin’
[ˈroboɾe] (for rōbŭr) > [ˈroβɾe] > [ˈroβle] roble ‘oak tree’

In a handful of cases, however, neuters in this category were remodelled as second-declension masculine nouns in /-o/:

capŭt > cabo ‘end’
vas > vaso ‘glass’
ŏs > hueso ‘bone’

 

4. References
Aebischer, Paul. 1971. ‘Le pl. -ās de la I^ère decl. latine et ses résultats dans les langues romanes.’ Zeitschrift für romanische Philologie 87:74-98.

Menéndez Pidal, Ramón. 1987. Manual de gramática histórica española (19th edition). Madrid: Espasa Calpe.

Penny, Ralph. 2002. A history of the Spanish language (2nd edition). Cambridge: Cambridge University Press.

Salvi, Giampalo. 2011. ‘Morphosyntactic persistence.’ In The Cambridge history of the Romance Languages Vol. 1, eds Martin Maiden, John Charles Smith and Adam Ledgeway (Cambridge: Cambridge University Press), pp. 318–81.