Secondary predication is poorly researched in Croatian linguistic literature. Among the few scientific papers on that topic there is only one monograph (Peti 1979). Peti’s monograph offers a consistent description of syntactic features of secondary predication while describing what Peti calls optional predicate extension. Monograph also provides a fairly detailed (though not complete) classification of morphological devices used to express secondary predication. But there are two problems with Peti's approach. First problem comes from the fact that Peti, in accordance with the early theory of generative grammar, insists on a rigid separation of syntax and semantics. Thus he ignores semantics for the most part, and relies on the autonomy of syntax principle. From that perspective Peti writes about a verb as a lexical, semantic unit which he then separates from the predicate as a formal, syntactic category. This in turn allows Peti to include in his description a category which he calls bound predicate extension. This category refers to predicate complements, i.e. predicate nominals selected by light or copulative verbs. By separating lexical features from predicate function, Peti maintains that predicate complements form predicates independent of light or copulative verbs which select them. In that way Peti believes that structures with light verbs and predicate complements contain two separate predicates in coordination. This is why he calls predicate complements extensions. Peti’s understanding of predicate complements is akin to a small clause theory in which predicate nominals are also viewed as predicates independent of the verb that selected them, and understood as predicates of a separate rudimentary clause. This doctoral thesis demonstrates that the assumptions upon which the categories of bound predicate extension and small clause theory are based are very problematic, especially when predicate nominals are tied to light and copulative verbs. Instead, we propose that in latter types of structures light or copulative verbs and predicate nominals form complex predicates. We define complex predicates as two separate predicate elements that function in synergy as one syntactic unit. Thus complex predicates are monoclausal structures. In contrast with complex predicates, structures with full verbs and adjunct predicate nominals are biclausal. Aside from the clause with full verb, secondary predicate can also function independently with controller as its subject and theme. However, the problem with Peti’s approach as with the small clause theory is that they do not explain how constructions with real secondary predicates occur. This becomes especially apparent when usage of full verbs is being modelled after the usage of light or copulative verb exemplars. Proponents of the small clause theory, for example, argue that the verb consider selects direct object and predicate nominal as a whole to which it then assigns a thematic role of proposition. But, if we take a full verb innovatively used after the usage of the verb consider as its model, to remain consistent with the theory of small clauses, we would have to analyse the direct object and the predicate nominal the same way as we would while analysing the verb consider. In that way, however, we would deprive full transitive verb of its direct object. Let’s imagine, for the sake of the argument, that the verb value is being modelled after the verb consider. So instead of a sentence like I consider him a fool we innovatively say I value him a fool. While the sentence I consider him a fool could be described as a small clause structure, e.g. [VP consider [SC him a fool], the sentence I value him a fool couldn’t, e.g. [VP value [SC him a fool] because the verb value would be left without direct object as its complement. In this doctoral thesis we demonstrate that the categories of bound predicate extension and small clause are problematic. Instead we try to explain the difference between complex and secondary predicates by relying on the theoretical framework and methods of construction grammar. We start with the assumption that in Croatian there is a predicative construction which comes in several semantic variants. We have defined and named these variants in accordance with semantic subevent expressed by specific predicate nominals found in them. Thus we have resultative, modal, similative, depictive and purposive or final predicate nominals. Each of these semantic variants of predicative construction is based on certain verb classes. These verb classes are in associative relations with that specific semantic variant of predicative construction. Collostructional analysis measures the strength of associative relation between individual verbs and specific variants of predicative construction under the term collostructional strength. Verbs that display the highest collostructional strength are verbs that elaborate the argument structure of a construction type. This also entails that they express a schematic, abstract meaning which makes them highly dependant on predicate nominals with which they form complex predicates. We maintain that secondary predicates exemplify cases of innovative integration of full verbs with various semantic types of predicative construction. These full verbs either belong to verb classes that are in associative relation with predicative construction although their collostructional strength is low or they belong to verb classes which typically aren’t associated with that specific construction type. Integration of a verb with various types of predicative constructions can be motivated by a specific verb serving as an exemplar, or it can simply be a result of a highly productive construction type in associative relation with a number of large verb classes whose exemplars are highly frequent in everyday language. An example of such a highly productive construction would be transitive construction which is being elaborated by a large number of semantically different verb classes. All the way from verbs of perception, e.g. to see a lion, and verbs of creation, e.g. to write a book, to verbs of possession, e.g. to have a problem. According to our understanding, secondary predicates are formed by integration of predicative construction with full verbs. In such examples we do indeed find two separate predicate elements in relation of coordination. One element is a verbal predicate and the other is a predicate nominal being controlled by a specific argument of whom it is being predicated. The second problem of Peti’s monograph is that it completely ignores the semantic side of secondary predicates. Description of semantic types of secondary and complex predicates is a fundamental goal of this doctoral thesis. Thus we describe five semantic types of complex and secondary predicates, i.e. five variants of predicative construction. Resultative construction or resultative complex and secondary predicates signify a change of state or a transformation that some referent has undergone. Similative construction or similative complex and secondary predicates signify a similarity between two referents mainly on the basis of the manner with which they are carrying out a specific action. Depictive construction or depictive complex and secondary predicates specify some feature or a function of a referent that is pertinent within an event denoted by a verb. Modal construction or modal complex and secondary predicates refer to subjective judgements by which speakers try to categorize or characterize various referents. Purposive or final construction describes a purpose or a function being assigned to a referent. Although we describe them as semantic types, each of the five types is also being described according to their syntactic features. This includes everything from describing their intransitive and transitive variants, argument structures as well as morphological features of predicate nominals that denote specific constructional subevents. Complex and secondary predicates in Croatian can be expressed by various morphological devices. Peti (1979) offered detailed (but not exhaustive) list. For example, his list does not include prepositional phrases with prepositions za ‘for’ and u ‘in’ that come with accusative complements as well as phrases with preposition do ‘to’ that come with genitive case. Although we will describe all these prepositional phrases in detail, our main material of investigation will be predicate nominals with predicate instrumental and word kao ‘as/like’. With the exception of resultative, predicate nominals with word kao can freely alternate with predicate nominals coded by predicate instrumental. The reverse, however, is not the case. Distribution of predicate instrumental in Croatian is noticeably more limited than distribution of predicate nominals in phrases with word kao. We believe that ever decreasing frequency of predicate instrumental in contemporary Croatian is a direct consequence of higher frequency of predicate nominals with word kao. Predicate nominals with word kao are suppressing predicate instrumental. There are examples of overgeneralization where speakers fuse both modes of expressing predication in one single sentence. In this thesis we will present several reasons why predicate nominals with word kao suppress predicate nominals coded by predicate instrumental. In general, aside from the two main types of secondary predicates, those being resultatives and depictives, there has been very little effort to describe various other semantic types of predicate nominals. This thesis therefore seeks to contribute to that specific underresearched area. Aside from that, by describing the processes of integration of full verbs with various types of predicative construction, our research contributes to understanding the relationship between adjunct predicate nominals and verbs. Rothstein (2006: 212) notes that the question of whether secondary predicates derive from syntactic or lexical operations is still open. We believe that the model of integration of full verbs with predicative construction presents a strong argument in favour of the view that secondary predication is a syntactic process. On the other hand we will show that complex predicates are formed by lexical operations.