This dissertation examines the classical problem of the metalinguistic word concept in the history of language studies. The problem is exposed under the perspective of Wittgenstein’s view of language, taking into consideration the works of Wittgensteinian linguists such as Roy Harris (1988, 1997, 2002, 2004), Sylvain Auroux (1992), Nigel Love (1989, 2004, 2007), Talbot Taylor (1996, 2000, 2008). The problem, then, is formulated in the context of the Radical Translation hypothesis of the analythical Philosopher W. Quine (1960). After that the following aim is set: to investigate the problem of the word as a linguistic and a category unit within the framework of the Cronstructionist Approach to language as proposed by Goldberg (1995, 2006), and extended by the typologyst W. Croft (1999, 2001). After presenting and exploring the theoretical assumptions, analytical principles and metalanguage used in the construcionist program, the recognition of the word as a unit and a category is evaluated within this language theory. It is demonstrated that the word has an ambiguous status in the constructionist approach: it hesitates between having its centrality subverted (replaced by the construction), and keeping the traditional Sausserean form/meaning pairing. Under the guidance of a Wittgensteinian perspective of language as a form of life a set of criteria is posed to indicate the theory which is best equipped to face the classical problem of the word. It is claimed that the Constructionist Approach, relieved from explicative and mentalist ambitions, is superior to the ones with which it contends due to the following reasons: (i) it is the one which best gives evidence of language as praxis, as a governed form of life, which is also irreducible and volatile, rather than a fixed and objective representational system; (ii) it is the one which best privileges description rather than explanation; (iii) it is the one that favors a holistic approach rather than an atomistic approach to language phenomena; (iv) it is the one that best accommodates the description of metalinguistic practices particular to specific speech communities – practices which are ultimately responsible for grammatical form –, being thus able to avoid an undesirable and yet historically recurrent metalinguistic imperialism.