Children with speech disorders often display difficulty in other domains of language, suggesting that they experience difficulty with the language learning process in general. A theoretical shift from viewing children's speech disorders as articulatory-based to viewing them from a linguistic perspective was precipitated by the application of phonological theories and principles to the field of speech pathology, beginning in the late 1970s. Implicit in this shift was the recognition that acquisition of a linguistic system is a gradual, primarily auditory-perceptually based process involving the development of receptive knowledge first (Hodson and Paden, 1991; Ingram, 1997). Another implication from research on normal phonological acquisition is that linguistic input should demonstrate a sound's contrastive role in the ambient phonology because what children are developing is phonological oppositions. As a result of applying a linguistic model to intervention, speech-language approaches have flourished in the past two decades for treating children's speech disorders. These approaches are characterized by: (1) an emphasis on the function of the phonological system to support communication, and thus on the pragmatic limitations of unintelligible speech; (2) a focus on the contrastive nature of phonemes and the use of minimal pair contrast training to facilitate reorganization of the system; (3) the use of phonological analysis and description to identify error patterns; (4) the selection of error patterns for elimination and of sound classes and features for acquisition; and (5) the use of a small set of sounds as exemplars of those patterns/features for acquisition. What separates speech-language approaches from other articulation approaches is the use of phonological analysis to identify error patterns affecting sound classes and sequences rather than the selection of isolated sounds to be trained in each word position. Speech-language approaches differ from articulation approaches in their focus on modifications of groups of sounds via a small set of exemplars (e.g., /f, s/ to represent all fricatives) and their emphasis on contrastivity for successful communication in a social context.

As such, minimal pair contrast treatment is a speech-language approach that highlights the semantic confusion caused when the child produces a sound error that results in a pair of homonyms (e.g., “sun” and “ton” both produced as /t∧n/). This technique involves contrasting a pair of words in which one word contains the child's error production and the other contains the target production (with phonemes differing by only one feature). In minimal pair approaches, the child is instructed to make perceptual and productive contrasts involving the target sound and his or her error. The goal of treatment is to help the child learn to produce the target sound in the word pair to signal a difference in meaning between the two words. Minimal pair contrast interventions have been shown to be effective in eliminating error patterns and increasing the accuracy of target and related error sounds (Ferrier and Davis, 1973; Elbert, Rockman, and Saltzman, 1980; Blache, Parsons, and Humphreys, 1981; Weiner, 1981; Elbert, 1983; Tyler, Edwards, and Saxman, 1987; Saben and Ingham, 1991). Minimal pair approaches that involve solely perception of word contrasts, however, have not resulted in as much change as when production practice with models and phonetic cues is also included (Saben and Ingham, 1991).

Minimal pair contrasts are used somewhat differently in Metaphon (Howell and Dean, 1994; Dean et al., 1995), a “cognitive-linguistic” approach. This approach is considered cognitive-linguistic because it facilitates conceptual development and cognitive reorganization of linguistic information. Its aim is to increase, at the metalinguistic level, awareness and understanding of sound class differences primarily through classification techniques, with little emphasis on production. For example, to contrast alveolars and velars, the concepts of front and back are introduced and applied first to nonspeech sorting and classification activities. Next, these concepts are transferred to the speech domain by having the child listen to minimal pairs and judge whether or not words begin with front or back sounds. Dean et al. (1995) present evidence from several children suggesting that this approach effectively reduced the application of selected phonological processes.

The cycles approach, proposed by Hodson and Paden's (1991), involves a goal attack strategy that capitalizes on observations concerning the gradual manner in which normally developing children acquire their phonological systems. Groups of sounds affected by an error pattern are introduced for only 1 or 2 weeks, then a new error pattern is introduced. Thus, the criterion for advancement to a new goal or target is time-based rather than accuracy-based. A cycle can range from 5 to 15 weeks, depending on the number of deficient patterns, and once completed, the sequence is recycled for the error patterns that still remain in the child's speech. Hodson and Paden's cycle approach involves auditory bombardment and production practice in the form of picture-and object-naming activities, and reportedly eliminates most of a child's phonological error patterns in 1–2 years of intervention (Tyler, Edwards, and Saxman, 1987; Hodson and Paden, 1991; Hodson, 1997).

In contrast to the three approaches just described, which focus on speech within a linguistic framework, there are language-based approaches in which little attention may be drawn to sound errors and these errors may not be specific targets of intervention. Instead, the entire language system (syntax, semantic, phonology, pragmatics) is targeted as a tool for communication, and improvements in phonology are expected from a process of “whole to part learning” (Norris and Hoffman, 1990; Hoffman, 1992). Phonological changes might be expected to occur because phonemes are practiced as parts of larger wholes within the script for an entire event. Language-based approaches involve a variety of naturalistic, conversationally embedded techniques such as scaffolding narratives, focused stimulation in the form of expansions and recasts, and elicited production devices such as forced-choice questions, cloze tasks, and preparatory sets.

Norris and Hoffman's (1990, 1993) language-based approach focuses on scaffolding narratives in the form of expansions, expatiations, and turn assistance devices to help the child talk about picture sequences with higher levels of discourse and semantic complexity. Hoffman, Norris, and Monjure (1990) contrasted their scaffolded narrative approach to a phonological process approach in two brothers with comparable phonological and language deficits. The narrative intervention facilitated gains in phonology that were similar to the phonological approach, and greater gains in syntactic, semantic, and pragmatic performance.

Other language-based approaches reported in the literature have focused primarily on morphosyntactic goals (e.g., finite morphemes, pronouns, complex sentences) using focused stimulation designed to provide multiple models of target morphosyntactic structures in a natural communicative context. Procedures have involved recasts and expansions of child utterances, and opportunities to use target forms in response to forced-choice questions, sentence fill-ins, requests for elaboration, or false assertions in pragmatically appropriate contexts (Cleave and Fey, 1997). Researchers have been interested in the cross-domain effects of these procedures on improvement in children's speech disorders. Fey et al. (1994) examined the effects of language intervention in 25 children with moderate to severe language and speech impairments who were randomly assigned to a clinician treatment group, a parent treatment group, or a delayed-treatment control group. The treatment groups made large gains in grammar after 5 months of intervention, but improvement in speech was no greater than that achieved by the control group. Tyler and Sandoval (1994) examined the effects of treatment focused only on speech, only on morphosyntax, and on both domains in six preschool children. The two children who received morphosyntactic intervention showed improvements in language but negligible improvement in phonology.

In contrast to these findings that language-based intervention focused on morphosyntax does not lead to gains in speech, Tyler, Lewis, Haskill, and Tolbert (2002) found that a morphosyntax intervention addressing finite morphemes led to improvement in speech in comparison to a control group. Tyler et al. (2002) investigated the efficacy and cross-domain effects of both a morphosyntax and a phonological intervention. Ten preschool children were assigned at random to an intervention of two 12-week blocks, beginning with either a block focused on speech first or a block focused on morphosyntax first. Treatment efficacy was evaluated after one block in the sequence was applied. Not only was the morphosyntax intervention effective in promoting change in morphemes marking tense and agreement in comparison to the no-treatment control group, but it led to improvement in speech that was similar to that achieved by the phonology intervention. Thus, for children who received language intervention, the amount of speech improvement was significantly greater than that observed for the control group. In a similar study, Matheny and Panagos (1978) examined the effect of highly structured interventions focused on syntax only and articulation only in children with deficits in both domains, compared with a control group. Each group made significant gains in the treated domain in comparison with the control group, but also made improvements in the untreated domain. Thus, a language-based intervention focused on complex sentence structures led to improved speech.

Findings regarding the effects of language intervention on speech are equivocal, particularly results from methodologically rigorous studies with control groups (Matheny and Panagos, 1978; Fey et al., 1994; Tyler et al., 2002). One variable that may account for these differing results is the use of different measures to document change. Matheny and Panagos used before and after standardized test scores, whereas Fey et al. used Percent of Consonants Correct (PCC; Shriberg and Kwiatkowski, 1982), a general measure of consonant accuracy, and Tyler et al. used a more discrete measure of target and generalization phoneme accuracy. Nonetheless, the collective findings from studies of the effects of different language-based approaches on speech suggest that some children, especially those with both speech and language impairments, will show improvement in speech when the intervention focuses on language. Determining exactly who these children are is difficult. Preliminary evidence suggests that children whose phonological systems are highly inconsistent may be good candidates for a language-based approach (Tyler, 2002). Finally, service delivery restrictions may dictate the use of language-based approaches in classroom or collaborative settings. These approaches deserve further investigation for their possible benefit in remediating both speech and language difficulties.

In summary, a variety of speech-language approaches have been shown to be effective in improving speech intelligibility and reducing the number and severity of error patterns in children with speech disorders. Although these approaches employ different teaching methods, they originate in a linguistic model and share an emphasis on the function of the phonological system to support communication, and on the contrastive nature of phonemes to reduce the pragmatic limitations of unintelligible speech.