Under pressure: Exploring the impact of cognitive factors on clitics placement in L2 Slovak

2.1 The principle of contiguity

The cognitive principle called ‘contiguity’ represents the concept according to which ideas, memories, and experiences are linked when one is frequently experienced with the other. There are two different types of contiguity: semantic and formal (positional). Semantic contiguity can be described as the closeness of two concepts which enables the conceptualizer to evoke one concept by the other. Formally, contiguity is realized as syntagmatic or morphological bondedness. The concept of bondedness is described by Lehmann as “the intimacy with which it [a sign] is connected with another sign to which it bears a syntagmatic relation,” and is understood as an indication of the degree of grammaticalization. “The degree of bondedness of a sign varies from juxtaposition to merger, in proportion to its degree of grammaticality” (Lehmann 2002, 131). In language theories, contiguity is described as the basis of syntagmatic relations (which are formed by contiguity) which oppose associate relations (which are formed by similarity) in Saussure’s synchronic two axes theory (Koch 1999, 142).

The cognitive preference for contiguity can be explained as being grounded in prominent efficiency-based accounts of word order which include various locality principles. Two locality principles in word order typology have been established in the theoretical studies. The first principle is dependency locality (DL), according to which languages tend to order words to reduce the overall distance between syntactically related words (Hawkins 1994). DL can be justified in terms of parsing efficiency, memory efficiency, and general communicative efficiency (Hawkins 1994; Gibson 1998; Hahn et al., 2020). The second locality principle is information locality (IL), whereby languages favour placing elements together when they are syntactically related or contextually informative about each other (Qian and Jaeger 2012).

Cliticization of formal elements representing certain grammemes or arguments of their matrix verbs and their discontinuous placement into the second position, distant from their matrix verb, can be considered as a violation of the DL principle as it includes the subordination of the clitic item under its prosodic host and (possible) disintegration from its syntactic or morphological governor. On the other hand, the verb-adjacent placement of clitic component minimizes total dependency length in accordance with the Dependency-Length Minimization hypothesis.

Research question 1: Do learners prefer those constituent orders which maintain adjacency of syntactically or morphologically bonded elements to enable short dependencies of the matrix verb and its clitic component in usage?

2.2 The principle of relevance

Speaking of the relevance principle, not only the proximity of language elements but also their constituent order should be taken into consideration. The principle of relevance can be ascribed to clitics components functioning within the inflection domain (verbal auxiliary clitics marking person and number in past tense and reflexive clitics marking voice in Slovak) and derivation domain (reflexive clitics functioning as derivational means for derivation of reflexive verbs).

With respect to grammatical markers, research by Bybee et al. (1990) scrutinized whether there is (i) a general preference for postposing grammatical material or (ii) a greater tendency to fuse elements that are already postposed. Their survey of a set of grammatical markers in a 71-language sample showed both.

Grammatical markers (Bybee et al. 1990, 5):

The research shows that postposed grammatical markers outnumbered preposed markers two to one (66%:34%), which explains the strong attraction of verbal auxiliary clitics as well as reflexive voice clitics to the right periphery of the matrix verb. The same tendency can be ascribed to reflexive clitics serving as valence markers (changing the number of arguments and valency of the reflexive verb derived from its non-reflexive counterpart) in Slovak (e.g. anticausative, deaccusative verbs). Their attraction to the position right to the matrix verb mirrors Universal Affix Ordering (Bybee 1985) according to which valence affixes succeed verb stem.

At the same time, both verbal auxiliary clitics and reflexive clitics in Slavic languages display grammaticalization processes from word to clitic and, in tendency, to postfix with the position on the right periphery after other derivational and inflectional suffixes. In a prototypical grammaticalization cline, according to Hopper and Traugott (2003, 7), clitic is understood as an intermediate stage between grammatical word and inflectional affix: content item > grammatical word > clitic > inflectional affix. However, this idea has been challenged by Haspelmath (2023), arguing that most types of clitics do not become affixes, and most types of affixes did not pass through a clitic stage. Especially ‘Wackernagel clitics’ (as found in Slovak) are unlikely to become affixes, as affixation requires a stable host whereas Wackernagel clitics are typical of promiscuity, i.e. word class selectivity. Postfixal status of certain types of clitics is evidenced in several Slavic languages (e.g. auxiliary postfix in past tense is evidenced in Polish, reflexive postfix in Russian). It confirms the tendency anticipated in research by Bybee et al. (1990) for postponed elements to be bonded. According to the principle of relevance (Bybee 1985), morphemes of greater relevance to the verb will generally occur closer to the verb stem; those of lesser relevance will generally occur at a greater distance. Reflexive clitics do not modify the verb semantics in a substantial way; the only change they bring into construction is to signal that the terminal point of the causal chain meaning is ascribed to the subject referent (in canonically reflexive verbs) or to trigger the implicature that the use of its lexical causative counterpart is too strong (in anticausative verbs).

These tendencies can be explained cognitively: the right peripheral position of inflectional formatives and derivation formatives with lesser relevance for the semantics of the verbal unit facilitates their processing through the effects of psychological primacy and recency and better outward indexicality towards the other parts of the sentence (Dressler et al. 1987, 7, cited in Haspelmath 1993, 292).

Research question 2: Do learners prefer those constituent orders in which clitic elements from inflection or derivation domains are placed on the right periphery of their matrix verbs?

2.3 The principle of verb–object bonding

According to the principle of verb–object bonding (Tomlin 1986), the relation between the verb and the object is stronger than that between the subject and the verb. The distribution of constituent orders across language families shows certain preferences with SOV being dominant, with SVO a close second, and VSO a distant third. The remaining orders (OSV, VOS, and OVS) are extremely rare (Dryer 2013). Furthermore, languages are known to change from SOV toward SVO, but the reverse change is much rarer (Gell-Mann and Ruhlen 2011).

There were several explanations offered for this fact, namely that SOV may be well-suited for describing non-reversible events, but it might be less well-suited for describing reversible events (Hall et al. 2013). Other explanations generally assume that what makes SOV problematic is the potential for confusability when two human nominals both precede the verb, which leads to role conflict (ibid.).

An alternative explanation for this issue is offered by the theory of uniform information density (UID). The UID hypothesis suggests, “language producers unconsciously endeavour to keep the rate of information transmission as close to constant as possible when speaking” (Maurits et al. 2010, 1587). In the light of the presented cognitive and pragmatic tendency, closeness to this UID ideal can explain the preferred distribution of certain word orders. According to this theory, object-first word orders deviate from uniformity because the first word (the object) carries a disproportionate amount of information as objects are predictive of very few subjects or verbs. For instance, the object word ‘water’ implies only a few possibilities for verbs (e.g. ‘drink’), which in turn restricts the subjects (e.g. to living things). By contrast, hearing the verb ‘drink’ implies many possibilities for objects (e.g. ‘water’, ‘coffee’, ‘cola’, ‘juice’, etc.; Maurits et al. 2010, 1592). This may explain verb–object order preference over object–verb order.

Research question 3: Do learners prefer constituent orders in which subjects precede verb and objects follow the verb?

3.1 Definition of clitics

Among factors determining the linearization of the components in the sentence structure in Slovak prosody, grammar and functional sentence perspective have been mentioned (e.g. Mistrík 1966). Informational structure, incarnated in the principles of functional sentence perspective, is considered the major principle of Slovak word order, whereas prosody and grammar play the role of accompanying factors as they are only employed when certain types of units appear in the sentence (namely clitic components with respect to prosody and attributives with respect to grammar).

Prosody relates to the word order realization of phonologically non-independent elements devoid of word stress called clitics which cannot be the sole elements of an utterance and cannot be realized freely, i.e., in different sentence positions depending on pragmatic and discourse function, but their position within sentence structure is determined phonologically.

However, phonological criteria like prosodic dependency or deficiency are not the most decisive features of these elements. As Haspelmath (2022, 20) states, clitics and affixes are very similar in that they are prosodically deficient, and there is no simple phonological criterion distinguishing affixes from words (including both free words and clitics). In many theoretical studies, two relevant characteristics of clitics are emphasized: bondedness, i.e. non-occurrence in isolation (which sets them apart from free words), and non-selectivity, i.e. lack of word-class selectivity or promiscuity (which differentiates them from affixes; Haspelmath 2022).

Slovak belongs to those languages which follow Wackernagel’s Law and its clitic elements belong to the category of second-position clitics (CL2). There are two features of 2P syntax: (a) that a combination of two (or more) phrasal categories X, Y preceding clitics in a CL2 language should be ungrammatical; (b) that the XP-position in any clause type with CL2 is not reserved for any particular syntactic category (e.g. noun phrase) and does not express any particular grammatical relation (e.g. subject; Zimmerling 2015).

3.2 Classification of clitics

In many theoretical works, unstressable and unstressed clitic components should be differentiated, the former labelled as clitics tantum and the latter as volatile clitics.

Clitics tantum can be defined as prosodically deficient unstressable elements that are always unstressed, independently of the context in which they are realized, thus being unable to be focused and moved into the initial position. Contrarily, volatile clitics can be defined as prosodically unstressed elements that can have phonological autonomy under certain contextual conditions. Clitics tantum which can only occur in the post-initial position is often referred to as ‘pure sentential clitics’ (Avgustinova and Oliva 1995) or ‘constant clitics’ (Rosen 2001, Hana 2007), whereas volatile clitics which may as well as not occupy post-initial position are labelled as semi-clitics (Avgustinova and Oliva 1995). A similar difference can be found in the work by Junghanns (2002) who differentiates lexical clitics for which clitic status represents an inherent part of their lexical ‘equipment’, and phonological clitics for which their clitic status is formed ‘in the phonological part of the grammar usage’ (elements that optionally undergo phonological cliticization). Functionally, clitics can be classified into several subtypes, depending on the grammatical and lexical status of the clitic component.

5.1 Errors with respect to type of clitic component and proficiency level

The first value to be investigated is the type of clitic component and the distribution of certain clitic types with respect to proficiency levels. Figure 1 displays the relative positions of both the row vectors – types of clitic components – and the column vectors – the proficiency levels.

Figure 1

Overall distribution of errors with respect to proficiency levels.

We can summarize the interpretation of the plot as follows:

1. conditional clitic is distinctively associated with C1,

2. clitic cluster and reflexive clitics are distinctively associated with B2,

3. copula is grouped with C2 (higher error ratio of non-auxiliary be-forms at C2 level is explained in section 6.4),

4. demonstrative is grouped with A1, and

5. B1 is associated with auxiliary.

To explain the attractions revealed by CA, the relative frequencies of clitic components in the texts from individual proficiency levels presented in Table 2 are compared with error rate at these levels.

The results suggest several meaningful discrepancies in lower versus higher proficiency levels concerning the ratio of particular types of clitics and their erroneous usage.

First, lower-level learners display an overall underuse of demonstrative clitics; however, erroneous usage of demonstrative clitics reaches the highs at A1. The possible explanation for this tendency lies in the fact that demonstrative items can be realized either as accented forms with corresponding freedom of occurrence, limited only by functional sentence perspective, or as unaccented clitics confined to second position. Two tendencies for erroneous placement of demonstratives are observable in the data.

1. Speakers often use non-focused argument demonstratives in postverbal position which can be explained by the principle of verb–object bonding:

   (9)	Určite	nevnímam	to	ako	pozitívum. (ru, C1)
   	certainly	neg-perceive-prs-1sg	it-acc	as	positive-acc
   	‘I certainly do not perceive it as a positive fact’
   	TH: Určite to nevnímam ako pozitívum

2. Speakers often place anaphoric demonstratives into initial position even in non-emphatic contexts which can be explained by the unstable status of demonstratives varying between full form and clitic and low pragmatic competence of speakers:

Second, learners use an increased amount of conditional clitics in intermediate levels, whereas erroneous usage of conditional marker by is overrepresented at advanced proficiency level (C1). This mismatch can be explained by two general factors:

1. higher error ratio of certain types of clitics can be linked to growing syntactic complexity of texts at higher proficiency levels which is reflected in more complex clause structures;

2. higher error ratio of certain types of clitics mirrors growing demands on information management reflected in word order preferences at prosodic–pragmatic interface.

The data show that erroneous usage of conditional clitics often occurs in sentence structures with multiple themes or with sentence adverbials functioning as disjuncts which strongly incline to be placed in the thematic zone of utterance (e.g. adverbials referring to source of knowledge) and may cause delayed placement of clitic component in the texts as communicative barriers:

These discrepancies between relative frequencies of clitic components in the texts and ratio of their erroneous usage in particular proficiency levels can be labelled as ‘frequency/error rate’ mismatch.

On the other hand, frequency distribution of past tense auxiliary clitics and their error ratio in particular proficiency levels show correspondence. Higher error ratio of auxiliary clitics at B1 level can be explained by overall distribution of past tense structures at this proficiency level.

5.2 Errors with respect to types of syntactic environments and proficiency levels

The second value to be investigated is the possible attraction between the erroneous usage of clitics components in certain syntactic environments with respect to proficiency levels. Figure 2 shows the relative positions of both the row vectors – types of syntactic environments – and the column vectors – the proficiency levels.

Figure 2

Overall distribution of types of syntactic environments with respect to proficiency levels.

We can summarize the interpretation of the plot as follows:

1. syntactic environments described as HVH, VH, and IS are grouped with C1 and B2;

2. syntactic environment described as JV is distinctively associated with A1.

The results point to an important fact concerning the level of syntactic complexity of texts at different proficiency levels. Erroneous usage of clitics in simple clauses is typical of low-level learners (A1) which may be caused by higher ratio of simple clauses in these levels, as the investigation for Slovak proved (Ivanová et al. 2021).

Growing syntactic competency leads to adhering more rigorously to word order rules in simple clauses at higher proficiency levels. On the other hand, more complex syntactic structures used at higher proficiency level attract more frequent difficulties with clitics placement. This tendency can be explained on the basis of the theory labelled as Barrier Theory which seeks to explain deviations from the basic order (2P in case of Slovak) as the result of different ‘barriers’ that hinder proper clitic placement. Barrier can be defined as certain type of sentence-initial groups or lexical heads hosting the clitic which have the properties (grammatical or communicative) due to which clitic element is moved from 2P to the right of this position (Zimmerling and Kosta 2013).

Grammatical barriers occur in certain types of syntactic environments (e.g. in main sentence with an interposed subordinate clause). In canonical word order templates, Slovak clitics stick to 2P position even when preceded by syntactic units which necessarily induce an Intonational Phrase boundary (Nespor and Vogel 1986). In texts produced by native speakers, clitics show default placement immediately after the IP boundary:

In such cases, clitics are prosodically parsed as proclitic on what follows them.

However, the data show that non-native speakers often employ delayed placement in these types of syntactic structures so that the clitic component is not placed immediately after the I-phrase boundary but after the second syntactic phrase:

Delayed placement occurs in cases when attributive subordinate clause splits the main clause so that the initial noun phrase with subordinate clause behaves as ‘bad clitic host’. Pancheva (2005) explains this pattern as a consequence of some need for clitics to be ‘second’ in their I-phrase.

The term ‘prosodic host’ should be differentiated from the term ‘anchor’, which is also used with respect to clitic components. According to Haspelmath (2022), anchor is the word preceding an enclitic, and the word following a proclitic, whereas host is the element with which a clitic forms prosodic word. Slovak clitics are prosodically neutral which means that they can lean prosodically to the following or preceding word depending on several factors. Prosodically, it is often determined by the length of the anchor. However, although prosodically neutral, clitics cannot follow just any pause but there are also syntactic factors relevant to the position of clitics. In those syntactic structures in which the subordinate clause precedes the main clause, the clitic is placed immediately after the I-phrase boundary only if the subordinate clause fulfils the role of the relative content clause:

The possible explanation for clitic ordering in main clause (mi nepovedal) is that the object (čo) of the complement predicate (nepovedal) is extracted from the main sentence and is realized as a complementizer introducing the subordinate clause, thus representing the first syntactic phrase after which the clitic component is placed in the expected 2P position. However, when no such extraction occurs, the subordinate clause is not parsed as an anchor for clitic and the clitic component thus occurs after the initial syntactic phrase of the main clause, as in example (18).

The identification of subordinate clauses with extraction and without extraction requires a high level of syntactic competence from the learners. The data show that the most common error in texts is the placement of clitic components immediately after the I-phrase boundary even in clauses with no extraction, as in example (19):

5.3 Errors with respect to distance of clitic component from its matrix verb and proficiency levels

The next parameter being analysed is the expected proximity of the clitic component from its matrix verb in terms of units labelled as linear dependency segments (LDS). Mačutek et al. (2021, 331) define the LDS as follows: “[…] the longest possible sequence of words (belonging to the same clause) in which all linear neighbours (i.e. words adjacent in a sentence) are also syntactic neighbours […].” It can be illustrated by the sentence from the data:

Four indices are calculated in the data annotation:

1. Expected distance between clitic and its matrix verb measured in the number of LDS (ED_LDS) and actual distance between clitic and its matrix verb measured in the number of LDS (AD_LDS);

2. Expected distance between the clitic and its matrix verb measured in the number of syntactic words (ED_SW) and the actual distance between the clitic and its matrix verb measured in the number of syntactic words (AD_SW).

In example (20), the expected distance of clitic component and its matrix verb with respect to number of LDS is 3 (TH: Tisícročné tradície sa [dnes] [už] [menej a menej odrážajú] v najmladšej generácii.), the actual distance is 2 (Tisícročné tradície [dnes] sa [už] [menej a menej odrážajú] v najmladšej generácii.), the expected distance from the matrix verb with respect to number of syntactic words is 4 ( = dnes, už, menej, menej), and the actual distance from the matrix verb with respect to number of syntactic words is 3 (už, menej, menej).

Discontinuous placement of the clitic component mostly relates to the ability of the learner to move it to the left sentence periphery independently of the position of the matrix verb. In the following figure, the expected and actual distance ratio is calculated, taking into consideration only the left sentence edge. Five distance values were taken into account: expected distance 0 LDS and actual distance 0 LDS (labelled as 0_0, the error is not caused by wrong distance placement in this case), expected distance 1 LDS and actual distance 0 LDS (labelled as 1_0), expected distance 2 LDS and actual distance less than 2 LDS (1 or 0, labelled as 2_10), expected distance 3 LDS and actual distance less than 3 LDS (2, 1 or 0, labelled as 3_210), and expected distance 4 LDS and actual distance less than 4 LDS (3, 2, 1, or 0, labelled as 4_3210).

Figure 3

Overall distribution of distance of clitic components from matrix verbs with respect to proficiency levels.

We can summarize the interpretation of the plot as follows:

1. A1 and A2 are mostly attracted to 0_0 distance (i.e. the syntactic structure of the sentence does not require non-adjacent placement);

2. the distance 3_210 is grouped with B1;

3. the distances 2_10 and 4_321 are distinctively associated with B2 and C1.

The data show that the growing syntactic complexity of the sentence structure calls for the distant placement of the clitic component on the left sentence periphery which often results in erroneous placement even at higher proficiency levels (B1, B2, and C1):

In example (18), the pronominal clitics ho is expected to be realized 4 LDS from its matrix verb (niekedy + treba + len + správne nasmerovať); however, its actual distance is 0 LDS.

5.4 Errors with respect to clitic placement on right or left periphery of the matrix verb

Following Zimmerling and Kosta (2013) and Zimmerling (2006), Slavic word-order systems can be classified into four types, tagged as ‘W’ for standard Wackernagel system with consistent 2P-clitics (Serbian, Croatian, Slovenian within Slavic languages in our data); ‘W⁺’ for modified Wackernagel system with verb adjacent clitics (Bulgarian within Slavic languages in our data); ‘W^*’ for degraded Wackernagel system with inconsistent 2P-clitics (Polish within Slavic languages in our data); and ‘C’ for communicative system (Russian, Ukrainian within Slavic languages in our data; Zimmerling and Kosta 2013, 201–203).

According to Franks and Holloway King (2000, 61), Bulgarian clitics are verb-adjacent rather than Wackernagel position clitics. However, pronominal clitics in Bulgarian behave comparably to other Sout Slavic and West Slavic languages, only higher Bulgarian clitics exhibit some idiosyncratic properties, especially clitic component li, which does not have an equivalent in Slovak. Therefore, Bulgarian is analysed as an instance of extended Wackernagel clitics system in this article. Similar behaviour is typical of Macedonian clitics; however, Macedonian learners of Slovak as L2 are not included in our data. East Slavic languages lack the kinds of pronominal and auxiliary clitics and the reflexive morpheme -sja became frozen as such (Franks and Holloway King 2000, 187).

On the basis of corpus data, the difference between Slavic languages belonging to standard, extended, and degraded W-systems and C-systems with respect to clitic placement on right or left periphery of the matrix verb is calculated. Three position variations are taken into consideration: 1) cases when the expected position of the clitic component is preverbal, yet the realized position is postverbal (PrePost); 2) cases when the expected position of the clitic component is postverbal, yet the realized position is preverbal (PostPre); and 3) cases when both expected and realized positions are preverbal (PrePre). In the latter case, the error arises as a consequence of the non-adequate distance of the clitic component from its matrix verb, not due to position on the left or the right periphery of the matrix work. The case when both expected and realized position is postverbal (PostPost) are not taken into consideration as there are only six instances of this type in the data.

Table 3 brings the percentual distribution of data in the group of L1 Slavic languages with W-systems and Slavic languages with C-systems.

The differences between frequency distribution within the group of Slavic learners from W-systems and C-systems were statistically tested by CA. The data show that there is no link between the rows and the columns of the table (as the computed p-value 0.341 is greater than the significance level alpha = 0.05, one cannot reject the null hypothesis H0: The rows and the columns of the table are independent). The lack of a significant difference between the two groups of Slavic learners is noteworthy. It suggests that Slavic learners, irrespectively of L1 classification of clitics into W-systems or C-systems, place the clitic components on left or right periphery of the matrix verb with a comparable error rate.

Therefore, the next parameter tested is the relation between the positional realization of clitics and the proficiency level. Figure 4 displays the relative positions of both the row vectors – clitic placement on the left or right periphery of the matrix verb – and the column vectors – the proficiency levels.

Figure 4

Overall distribution of types of clitic placement with respect to proficiency levels.

In the case of clitic placement with respect to the right or left periphery of the matrix verb, the data show significant differences between the proficiency levels. These findings suggest that clitic placement with respect to matrix verbs undergoes syntactic development.

We can summarize the interpretation of the plot as follows:

1. PostPre position is grouped with B1;

2. PrePost realization is associated with A2;

3. PrePre shows attraction to C2.

The development of word-order competence shows different development trajectories depending on the proficiency levels. Early stages (mostly A2) are typical of erroneous placement of clitic components into postverbal position, i.e. on the right periphery of the matrix verb (PrePost), irrespective of the prosodic pattern of the utterance which calls for placement of clitics on the left periphery of the matrix verb.

Another possible explanation of erroneous clitic placement in the first sentence is the role of parallelism: first, grammatical parallelism (i.e. clitic components in a compound sentence are both the auxiliary clitics), and second, positional parallelism, with respect to the position (the position of clitic in the second sentence influences the placement of clitic on the right periphery of the matrix verb in the first sentence).

The following stage (mostly B1) is characteristic of errors concerning placement of clitics into preverbal position instead of postverbal position (PostPre). It can be explained on the basis of development of syntactic complexity in the interlanguage. Numerous theoretical studies point to the fact that syntactic complexity in L2 is thought to expand from coordination to subordination and then to phrasal elaboration, as learners gain proficiency (Wolfe‐Quintero et al. 1998). At beginner and low–intermediate proficiency levels, syntactic growth may show an increase in coordination (Bardovi‐Harlig 1992, Vyatkina 2012) and upper–intermediate levels are thought to display an increase in subordinate structures. The same results have been proved for Slovak as L2 (Ivanová et al. 2021). Erroneous placement of clitic components into preverbal position occurs mostly in compound sentences whose ratio ascends at A2 and B1 proficiency levels.

In Slovak, elements such as a ‘and’ cannot serve as prosodic hosts. The same holds for Bulgarian i ‘and’ which nonetheless obviate the reordering phenomenon, e.g.

In principle, clitics in Bulgarian are preverbal elements. When there is no nonverbal lexical material that could support clitic placement in front of the verb, the verb itself must precede the clitics to provide phonological support for them. This reordering strategy is known as the Tobbler-Mussafia effect. However, even i ‘and’ which cannot serve as a prosodic host is sufficient to serve this purpose to keep the clitic component in preverbal position and thus obviate TM to occur (Franks 2008, 99).

In the data, 50% of the errors occurring in this type of syntactic construction was produced by Bulgarian speakers, e.g.

Finally, the last stage of interlanguage development (advanced proficiency levels, mostly C1 and C2) is connected with erroneous placement of clitics concerning their distance from the matrix verb (PrePre). At this stage, learners are able to put the clitic component in the position on the left periphery of the matrix verb, depending on prosodic circumstances in the sentence; however, the errors mostly occur as the result of inappropriate distance of the clitic component from the matrix verb.

The next parameter that was tested was the relation between the positional realization of clitics and the type of clitic component. Figure 5 displays the relative positions of both the row vectors – clitic placement on the left or right periphery of the matrix verb – and the column vectors – the type of clitic component.

Figure 5

Overall distribution of clitic placement on the left or right periphery with respect to type of clitic component.

In the case of clitic placement with respect to the right or left periphery of the matrix verb, the data show significant differences between the types of clitic components. The CA provides two significant attractions:

1. PrePre position is grouped with copular clitic;

2. PostPre realization is distinctively associated with demonstrative clitic.

The data show a strong preference for adjacent realization of clitic and its predicative complement in the interlanguage of learners. The expected distance of clitic be-form from its predicative complement is 1.06 LDSs and 1.6 syntactic words, yet the actual distance ratio is 0.06 LDSs and 0.07 syntactic words which means that the learners tend to place clitic be-form adjacently to its predicative complement irrespective of its clitic nature.

The possible explanation for errors in usage clitic be-forms is the nature of the element which occupies the 2P position instead of clitic. According to the data, this element is predominantly subject (49%) or adverbial (27%), i.e. the components which fulfil the role of thematic elements, namely the role of Bearer of Quality (B) as in example (23) or (locational or temporal) Setting (Set) as in example (24):

In these cases, the clitic be-form undergoes delayed placement which is caused by a communicative barrier (forwarding theme element).

Cognitively, the preferential placement of be-form in an adjacent position to its predicative complement can be understood as a trade-off between information marking and word order of clitics. It is connected to the idea that communication is a noisy phenomenon and communicators strive to minimize the potential for errors on the side of receivers (noisy channel account; Gibson et al. 2013). In be-constructions, both predicative and subject nouns can be used in nominative case in Slovak.

Within the structure in which subject and predicative noun are located adjacently as in the target hypothesis in example (25) (ak je homosexual veriaci), the role difference between subject or predicative noun is lost due to noise which element is the subject and which element is the predicative noun. Interposing clitic components between two nouns is more robust to noise because even when information about one sentence component is cancelled out; the interlocutor can tell solely from the position with respect to the verb whether the information concerns subject or predicate noun: ak homosexuál je = subject, ak je veriaci = predicate noun. This situation holds for reversible structures (when both subject and predicate noun are semantically homogenous and both components can fulfil the role of subject and predicate noun).

The data also show the preferential placement of demonstratives into preverbal positions. It can be explained by the semi-clitic status of demonstrative components which under certain circumstances may occupy the initial position.

Demonstrative pronoun functions as a clitic component in structures in which it refers to a topicalized antecedent noun phrase. In these constructions, the reference can be viewed as a process of elaboration (the noun phrase is being elaborated on or made more accurate) with the classifying or identificational adjectives modifying predicative complement. The canonical position of demonstrative elements behaving as volatile clitics is 2P in these types of structures. On the other hand, non-clitic realization is preferred for cases when the clitics resume a whole sentence or text passage and the reference can be seen not as elaborating but as evolving (it signals a new denomination, often expressing also evaluative attitude).

The differentiation of reference as elaborating or evolving requires a high level of pragmatic and syntactic competence from the learners. The data show that the most common error in texts is the placement of demonstrative clitic components into initial position in clauses with elaborating references as in example (26) or cliticization of demonstrative elements in clauses with evolving references as in example (27):

Presented word-order patterns differ with respect to index referred to as ‘wdist’ (distance between anaphor and antecedent in words; Strube et al. 2002), which is usually described as one of the coreference-level features contributing to successful anaphor resolution (linking the referring expression to its antecedent). Within Accessibility Theory (Ariel 2001) anaphor resolution is linked to the appropriate degree of cognitive accessibility of the antecedent referent. Accessibility depends on several factors: the antecedent’s salience, its recency of mention, and cohesion between clauses containing antecedent and anaphor, etc. Clitic usage of demonstrative pronoun posits wdist 1 in clauses with elaborating reference, whereas its non-clitic usage leads to wdist 0 in clauses with evolving reference.

In example (26), the antecedent noun phrase najvyššiu a najmladšiu plošinu has a high level of salience – it has been shown that noun phrases containing attributes as well as referents being the goal of action (nájdeme plošinu) are perceived as more prominent than noun phrases without attributes or with other semantic roles (Klin et al. 2004, Arnold 2001). High-salience antecedents can have their referring expressions (i.e. demonstrative clitic) more distantly than low-salience referring expressions.

On the other hand, the demonstrative element in example (27) resumes the whole previous sentence passage which as an anaphoric non-nominal antecedent has a lower degree of salience. This fact is thus balanced by zero distance of anaphoric demonstrative from its referring expression, resulting into its non-clitic status. Thus, each position realization of demonstrative can be interpreted as referring to a referent of a particular salience level. Non-clitic use of a demonstrative signals a less salient antecedent referent whereas a more salient antecedent referent admits clitic usage of demonstrative.