The theory of generative grammar in linguistics is pursuing scientific explanation by setting up phrase structure rules to derive deep structure strings which are transformed into surface structure utterences. The whole process involves the deductive-nomological model. There are two approaches to semantic analysis for explanation: logical and psychological. The logical position of semantics aims at the characterization of the notions of truth of a sentence relative to a model and of entailment between sentences. It is concerned with formal precision, consistency and elegance, not with psychological processes. The psychological position, on the other hand, tries to treat meaning in terms of all aspects of cognition. For the most satisfactory explanation of semantic phenomena, both positions are necessary and must be synthesized or integrated. Major areas of linguistic meaning involve both logical and psychological investigations. The treatment of durational expressions with the universal quantifier is insufficient and there must be consideration of how we perceive the reality and of the perceived relevant psychological time. Presupposition is an area in which psychological investigation challenged the two-valued logic and made the notion of truth-value gap possible. Emotives are stronger in factive presupposition than epistemic verbs such as ``discover,`` ``realize``. In Korean the verb ``al`` (=know) is used with or without factive presupposition of the complement depending on whether the complement is followed by the object marker (then presupposed) or the directional marker. Different degrees of presupposition cannot be ignored in semantic investigation. The intensions of basic lexical items are assumed as givens in Montague``s semantic rules. But the meaning of a lexical item crucially involves the psycholoical process of categorization. Even Putnam proposes the notion of a ``steredtype`` of a natural kind term. Rosch``s investigation shows that categories, particularly basic level categories, are not arbitrary and ate ihternally structured by gradients of representativenss``. In logical semantics, meaning postulates are used to show implication between nonlogical lexical items. Catnap originally did not think of cross-linguistic concepts involved in atomic predicates of meaning postulates and did not consider factual knowledge. However, for the meaning postulate method to be useful, it must have an empirical basis and take into account the psychological process of categorization. The problem of propositional attitude sentences also calls for psychological explanation. Formal semantics has not been quite successful in determining the truth conditions of propositional attitude sentences. There are some recent approaches to this problem and they involve certain psychological aspects. Psychological constraints are needed to characterize the semantic theory of human languages. Linguistic semantics as an empirical science calls for scientific explanation, not just a self-contained formal system. The ``mentalism`` advocated by generative grammarians purports to get interested in theory and explanation rather than in discovery procedures and taxonomy. It is not the mentalism in the dualistic sense of the term.