Introduction

In a broad sense, a child whose speech is coherent and complex can be considered a fluent speaker. When a child speaks fluently, the listener has the impression that the psycholinguistic processes of speech planning and speech production are functioning easily and efficiently (Lennon, 1990). Whereas fluent speech is an ‘automatic procedural skill that does not require much attention or effort from the speaker’ (Schmidt, 1992: 358), non-fluent speech requires substantial effort and attention. As a consequence, the speech of non-fluent speakers exhibits a considerable number of hesitations and other manifestations of groping for words as well as unsuccessful attempts to combine them into utterances.

Self-repairs: Self-monitoring versus metalinguistic competence

There is no agreement in the literature as how to label the set of pauses, repetitions, revisions and abandoned utterances that characterize a non-fluent speaker. This lack of consensus often responds to different interpretations of the functions, purposes and processes associated with disfluencies (see MacLurg, 2014, for a review). For instance, Postma et al. (1990) propose a classification of these processes on the basis of three criteria: (i) speech errors or deviations of speech planning (e.g. phoneme substitutions); (ii) disfluencies or disruptions of speech planning (e.g. pauses, repetitions, sound lengthening); and (iii) self-repairs, which represent the knowledge a speaker has of their linguistic system to fix errors (self-monitoring). Children's self-repairs are based in their capacity to monitor their role in the conversation, adjust their speech acts and remediate potential problems or mismatches between the child's and the adult's linguistic systems (Morgenstern et al., 2013).

In the present study, we adopt the taxonomy proposed by Fletcher (1990), which in turn is based on Levelt's (1989) model of speech production.