The conquest of the Iberian Peninsula by Romans was followed by the spread of the Latin language on the territory of Spain. It was the expression of territorial and cultural unification of the people of the Roman Empire.

The Romanization of the Iberian Peninsula lasted approximately two centuries. The Latin language was so deeply rooted in Spain, that local languages (except the Basque language) had almost disappeared. However, even nowadays the trace of the pre-Roman substratum is noticed in the Castilian language, while the Spanish phonetics is greatly influenced by the pre-Roman languages of the peninsula.

In addition to the influence of the pre-Roman languages, the Latin language (which was spread in Spain) acquired the German and Arabic elements during the following stages of its evolution. Its original form was changed and the foundation of the Spanish Romance was facilitated.

Despite the fact, that the Spanish language was closely related to substrate (Iberian), superstratum (West Gothic) and adstrate (Arabic) languages, a lot of Spanish lexical units have the Latin origin. Therefore, the majority of binary and poly-constant complexes are connected to the Latin roots. Hence, in some cases, consonant complexes are found in the words of the pre-Roman, Greek, German and Arabic origin.

The given paper deals with the study of consonant groups - two-element complexes. It describes the evolution of the groups and the reasons causing their variations.

Our research was based on the data of the newest etymological dictionaries of Corominas, Garcia de Diego and Santiago Segura Mungia (also the Royal Spanish Academy - RAE). During the research of the material we found the consonant complexes with a high degree of sustainability and the groups, which had underwent single or several phonetic changes during the process of their evolution. We also discovered the consonant complexes, which are not met on the level of etymon of the Latin, Greek, pre-Roman, Arabic or other languages. However, they are found in the Spanish correlate.

On the basis of the gathered material we have grouped the studied consonant complexes in the following manner: 1.Consonant groups, which have a high degree of stability; 2. Consonant groups with a tendency of variation; 3. Consonant complexes, which enter the Spanish language with unchangeable or changeable forms; 4. Consonant complexes, which exist on the level of a derivate. At the same time, we determined the phonetic changes, which conditioned the variation of complexes.

Similarly to the group bt, the middle consonant complex bs can be distinguished by its high degree of stability. Neither bt, nor bs undergo phonetic changes in the process of evolution and enter the Spanish language with unchangeable forms: Latin observator → Spanish observador, Latin obturare → Spanish obturar, etc.

The consonant complex dr is usually met in the words of the Latin origin. It transforms into the Spanish correlate without variation: Latin cuadrigalis → Spanish cuadriga, Latin cuadrus → Spanish cuadro.

The group fr is not subject to the variation. It is found only in the words of the Latin etymon and is distinguished by a high degree of sustainability, for instance, the Latin afrontare corresponds to the Spanish afrontar.

Five binary complexes beginning with the consonant l transform into Spanish in  unchangeable forms. The consonant groups ld, lf, lg, lm and lv are found in the words of the Latin and non-Latin origin: Latin caldus → Spanish caldo, Latin alga → Spanish alga, Latin palma → Spanish palma, Latin salvus →  Spanish salvo.

It's interesting, how the complex lf paved the way in the Spanish alfil. It had derived from the Persian pil, which added an article in Arabic and finally formed al-fil. It's clear, that this consonant group emerged in Arabic (on the level of derivate) and unchangeably transformed into Spanish. The same occurred during the etymological research of the Spanish alfañique, which derived from the Persian panid via adding the article in Arabic (al-fenid).

The consonant complex nv can be distinguished as one of the most stable complexes. The group composed of a sonorous nasal and a sonorous fricative is mainly found in the verbs, which are formed via adding the prefixes con and in to the root-words, for instance, Latin convergere (con+vergere) -  Spanish  convergir, Latin  invadere (in+vadere) - Spanish invadir. In certain cases nv complex is a root. Therefore, it is not created by means of two morphemes, for example, the Spanish convexo and invocar derived from the Latin convexus and invocare [Segura Munguía, 2010: 889].

In contrast to the initial position, the consonant group pl is not subject to variations in the middle of the word: Latin triplex → Spanish triple, Latin triplicare → Spanish triplicar, etc. The same can be said about the group pn, for instance, the Greek hypnos (sueño) becomes the Spanish hipnosis. As it turns out, the Greek stem is maintained in the Spanish derivative, while the addition of the Spanish suffix "sis" indicates to the process.

In the list of the most spread and stable consonant complexes an important place is occupied by the groups having r as an initial element. Therefore, the groups rb, rl, rm, rn, rr, rv  are found in the words of the Latin, Greek, Celtic, Gothic and Arabic origin. Their variations are not met in the Spanish correlates.

The Spanish arbol has the Latin etymon. It derives from arbor. It's obvious, that the consonant complex of our interest is maintained. The same cannot be said about the last consonant, which transforms after the dissimilation. The Spanish berlina has two semantic meanings. When berlina denotes "a car with four doors", it derives from the French berline, which contains a vocable established after the name of Berlin (the place, where it was invented). If berlina means "mocking at someone", it has the German origin and transforms into Italian as berlina (jester) [Coromines, 2008: 73]. The Greek harmonia entered into Spanish as armonía. The most words with rn complexes have the Latin etymon: pierna (Latin. perna), cuerno (Latin. cornu). In contrast to the consonant group rn, the vowels e and o had undergone a phonetic variation. As a result of the diphthongization, the vowel e transformed into ie, while o became ue. The geminate of r is met in Latin and in the words of the pre-Roman etymon. The Spanish berrendo has the pre-Roman (Celtic) origin and comes from barrovindos, whereas the Latin arrogans forms the Castilian arrogante. In the list of stable complexes beginning with the consonant r, the complex rv occupies the last position and is presented only in the words of the Latin origin: Latin cervus → Spanish ciervo, Latin corvus → Spanish  cuervo, Latin  parva →  Spanish parva ,etc.

There is a significantly small number of consonant complexes with a tendency to the variation. During the research we discovered only three groups. All of them are geminates. The voiced occlusive geminate bb as well as the voiceless occlusive geminate pp undergoes variation and enters the Spanish language in a simplified form. The simplification of the double occlusive is shown in the following words: Latin iabbatia → Spanish abadia, Latin abbatissa →  Spanish abadesa, Latin cappa → Spanish capa, Latin appetitus → Spanish apetito, etc. It's clear, that in the first two examples the voiceless consonants were voiced in addition to the formation of the monophonemic correlate. The geminate m was simplified in the following Spanish words: comunicar (Latin comunicare), comunidad (Latin communitas). Moreover, we found a variation of the same geminate in which the first element of the complex transforms into n: conmoción (Latin commotio), conmutar (Latin  commutare), etc.

In the third group we have united consonant complexes, which transform into Spanish without variations or with changed forms. This group comprises the majority of the consonant groups.

The consonant complex gl is a rare phenomenon. Hence, there are examples of its variation and constancy, for instance, the Latin degluttio transforms into the Spanish deglución, while the Latin espeglo becomes the Spanish esplego. It's obvious, that in the second example the removal of l decomposes the consonant complex gl and creates monophonemic correlate of the Spanish language.

The loss of the first element explains the simplified transition of the consonant complex gm into the Spanish language: Latin augmentum → Spanish aumento, Latin pigmentum → Spanish pimiento.  Its constancy is presented in the Spanish word segmento, which derived from the Latin segmentum [Real Academia].

Phonetic changes of the consonant complex gn can be divided into three subgroups. The derivation of the monophonemic correlate from a binary group (on the expense of the lost first element) is presented in the Spanish conocer (Latin cognoscere), while the palatalization of the complex is given in leña (Latin ligna) and estaño (Latin stagnu) [Penny, 2008: 90]. The Spanish prenda had a difficult evolutional way. It originated from the Latin pignora and after several changes formed the Castilian correlate:  pignira → peñra → peñdra → prenda. It's obvious, that the consonant complex of the Spanish correlate, which existed at the level of the etymon disappeared after the palatalization and epenthesis or metathesis of the consonant. However, an unchangeable form of gn moved to the following Spanish words: magnífico (Latin magnificus), magno (Latin magnus), etc.

Despite the fact, that the middle consonant complex ln is not subject to the variation (Latin balnearius → Spanish balneario, Latin alnus → Spanish alno), there are two examples which contradict this fact. After the palatalization, l disappears and the change of the degree of the last consonant gives the palatalized n or ñ: balneum → bannium → baño [Akhvlediani, 2009: 67].

The consonant group lp is found in the words of the Latin and non-Latin origin. There are several examples of its variation and constancy. During the establishment in the Castilian language, the lp complex maintained its initial structure in the following words: Latin palpitare → Spanish palpitar, Latin pulpitum → Spanish pulpito, etc. The vocalization of the first element was presented during the evolution of the Latin talpus. Its first version was taupo, but afterwards, the diphthong au underwent monophthongization and today's Spanish uses topo. The fact of dissimilation explains the derivation of parpadro and afterwards, párpado from the Latin palpetrum.

One of the most spread consonant complexes is mb group. Despite the fact, that there are numerous examples of its sustainability (Latin   ambitus → Spanish ámbito, Latin ambulare → Spanish  ambular), we come across the simplification of this complex on the level of the Spanish correlate, for instance, the Latin palumbes (palumba) is simplified under the Gothic influence and forms paloma [Akhvlediani... 2009: 42]. Moreover, the Latin umbilicus forms ombligo via the syncope of i, which exists between b and l.

The tendency of mn complex towards assimilation was noticed in Latin. At first, the Latin damnum was modified at the etymon level. Afterwards, the assimilated dannu transformed into daño. It's obvious, that the geminate of n (formed in Latin) moved to Spanish as palatalized n (or ñ), while the consonant group mn moved into Spanish omnipotente without variation. It's also worth mentioning, that omnipotente originated from the Latin omnipotens.

The discussed material has shown, that ng complex is one of those groups, which are often found in the Spanish correlate in a modified form (seldom - without variation). Its phonetic changes can be divided into several groups. The first group unites the binary complexes which became monophonemic at the Spanish correlate level. This process was preceded by the palatalization and change of the "degree" of the last consonant (Latin  cignere → Spanish  ceñir). The second group encompasses examples with the lost second element (Latin pungente →  Spanish puniente). The third group embraces examples of the transformation of binary complexes into poly-consonant ones (the Latin conger transformed into the Spanish congrio via epenthesis of r). The fourth group includes examples in which the second element is transformed. This group can be divided into two subgroups: a) words in which g transforms into c (singellus-sencillo); b) words in which the second element becomes ch: cingulum-cincho. The ng complex was unchangeably transformed into the following words: angina, angor, largo, etc.

Several cases of variation were found during the study of the middle consonant nt. There were examples of its constancy as well, for instance, nt was unchangeably transformed from Latin into the Spanish contacto, intentar, etc. During the process of variation we found the voicing of the voiceless element of the two-element complex (antubiya → Spanish endivia) and the  transformation of the voiceless consonant into c (Latin amentia → Spanish amencia) or into z after a, o and u (Latin pantex → Spanish panza) conditioned by the necessity of correct pronunciation. Moreover, the transformation of a binary complex into a poly-consonant one was found in the Spanish entraña, which derived from the Latin interaneus.

The consonant complex pt is found only in the words of the Latin origin. It mainly transforms into the Spanish correlative in a modified manner. We have classified these alternations into several groups. The first group unites words with a monophonemic correlative, which  emerged on the basis of the loss of the first element (Latin ruptu → Spanish roto). The second group deals with the words with the excluded binary groups (Latin  septimana → Spanish semana). The third group includes the examples of vocalization (Latin captivus → Spanish  cautivo). The pt complex was unchangeably transformed into several words, for example, captura (Latin captura), etc.

The middle consonant complex rd is presented in the words of the Latin, Greek and pre-Roman origin. It mostly transforms into Spanish in an unchangeable form: Latin ardere → Spanish arder, Greek kardiakos → Spanish cardiaco, Celtic bardus → Spanish bardo. There are only two cases of the variation of rd complex: the Latin hordeum formed the Spanish orzuelo (presumably, under the influence of the Italian orzata) and horchata [Coromines, 2008: 302].

The consonant complex rp is mainly transformed in an unchangeable form: Latin corpus → Spanish cuerpo, Latin serpente → Spanish serpiente. There is only one case of its variation: the verb sobreponer, which derived from the Latin superponer. Therefore, the metathesis and voicing of the voiceless consonant p are noticed at the derivative level.

One of the most widespread forms of assimilation is the consonant complex rs. It was presented in some words of the ancient Latin. Afterwards, rs changed into ss. Moreover, after the 18^th century the simplification of the geminate of the given words has been carried out on the expense of the omission of one of the elements: transversu → traviesso → travieso, ursu → osso → oso, etc [Akhvlediani, 2008: 70].The rs complex was unchangeably transformed into several words of the Greek origin, which were presented in the Spanish language in the Latinized form: Greek thyrsos  → Latin thyrsus → Spanish torso.

Rt complex underwent different changes during its establishment in the Spanish language. Besides the fact, that the second element was voiced (Latin fartum → Spanish fardo), the transformation of the voiceless occlusive into the voiceless fricative was presented (Latin martialis → Spanish marcial). There were also the cases of transformation of the second element into the voiceless affricative (Latin cortex → Spanish corcho). Hence, the consonant complex rt transformed unchangeably into the following Spanish words: corto (Latin curtus), arteria (Greek arteria), etc.

The rt complex rarely undergoes changes on the level of etymon. However, there are several cases of its variation. Among them is the Spanish cidro, which derived from the Latin citrus via voicing of the consonant. The consonant group rt transformed into the Spanish atrio and matraca without variation. The former originated from the Latin atrium, while the latter derived from the Arabic matraqa.

The fourth group unites consonant complexes, which are presented on the derivative level. The groups created on this level are often called "grupos secundarios" [Morales, 2005: 217]. Hence, they are small in number. Therefore, we included in the fourth group other complexes, which are found at the level of etymon of some words and at the derivative level of certain vocables. For this reason, we divided the given group into two subgroups. The first group includes sf, sg middle consonant complexes, while the second group unites dr, gl, gm, gn, ld, lg, lm, lp, lv, mb,  nt, nv, rf, rg, rl, rm, rn, rp, rr consonant groups. It's worth mentioning, that the words of the first subgroup are vocables of the non-Latin origin: Greek sphaera → Spanish  esfera, Basque izokin → Spanish esguin. It's obvious, that in the first example the complex existing at the level of etymon transformed phonetically. Moreover, it changed the position in the word.

The formation of the consonant complexes of the second subgroup in the Spanish correlate is connected with the syncope, metathesis, assimilation, dissimilation and voicing of the voiceless consonant.

Gl and rm consonant complexes are formed as a result of syncope: Latin regula → Spanish regla, old German aram → Spanish armón. There are several cases when the syncope of a vowel is accompanied by the voicing of a voiceless consonant. The Spanish groups dr, ld, lg were formed in this way, for instance: Latin lateralis → Spanish adral, Latin malitas → Spanish  maldad, Latin delicatus → Spanish delgado. In several words the same dr complex is derived from the metathesis: Latin crocodilus → Spanish cocodrilo, Latin brodium → Spanish bodrio. In the Spanish word párpado the rp complex is formed via dissimilation. This word originated from the Latin palpetrum, which transformed into parpadro and created párpado. The assimilation is presented in the Arabic words beginning with r. They usually add the article al in the process of their establishment in the Spanish language. L of the given article produces a geminate of r after the assimilation with r of the root, for instance, Arabic ruzz → Spanish arroz.

The discussed material has revealed, that the Latin and non-Latin consonant complexes passed a difficult way of evolution in the process of their incorporation into the Spanish language. The majority of phonetic changes were caused by linguistic (internal) factors connected with the articulatory apparatus of communication. The voicing of voiceless consonants (also, in several cases of assimilation, dissimilation, syncope and epenthesis) in the intervocal position can be regarded as a clear example of this fact. Moreover, the adoption of words from Latin and other languages in unchangeable forms was unacceptable for the people inhabiting the Iberian Peninsula. For this reason, many phonetic (systemic and non-systemic) changes were presented in these words.

During the research we came across the variations connected with extralinguistic factors. The acoustical error - named by Menéndez Pidal as "error de audición"   [Menéndez Pidal, 1977: 195] - explains the variation of several consonant complexes. There are different types of mixtures, but the "error" of the point of articulation is found between occlusives and nasals in cases of consonant groups: bieldo →  bielgo, tunbak  → tumbaga.

It's obvious, that the Spanish language has undergone structural changes (hence, it maintained inherited elements). Its phonetic system is not less difficult than the Latin one.