In evaluating the properties of bilingual speech, an anterior question that must be answered is who qualifies as bilingual. Scholars have struggled with this question for decades. Bilingualism defies delimitation and is open to a variety of descriptions and interpretations. For example, Bloomfield (1933) required native-like control of two languages, while Weinreich (1968) and Mackey (1970) considered as bilingual an individual who alternately used two languages. Beatens-Beardsmore (1982), observing a wide range of variations in different contexts, concluded that it is not possible to formulate a single neat definition, and stated that bilingualism as a concept has “open-ended semantics.”

It has long been recognized that bilingual individuals form a heterogeneous population in that their abilities in their two languages are not uniform. Although some bilingual speakers may have attained a native-like production in each language, the great majority are not balanced between the two languages. The result is interference from the dominant language. Whether a child becomes bilingual simultaneously (two languages are acquired simultaneously) or successively (one language, generally the home language, is acquired earlier, and the other language is acquired, for example, when the child goes to school), it is impossible to rule out interference. In the former case, this may happen because of different degrees of exposure to the two languages; in the latter, the earlier acquired language may put its imprint on the one acquired later. More children become bilingual successively, and the influence of one language on the other is more evident. Yet even here there is no uniformity among speakers, and the range of interference from the dominant language forms a continuum.

The different patterns that a bilingual child reveals in speech may not necessarily be the result of interference. Bilingual children, like their monolingual counterparts, may suffer speech and/or language disorders. Thus, when children who grow up with more than one language produce patterns that are erroneous with respect to the speech of monolingual speakers, it is crucial to determine whether these nonconforming patterns are due to the influence of the child's other language or are indications of a speech-language disorder.

To be able to make accurate diagnoses, speech-language pathologists must use information from interference patterns, normal and disordered phonological development in general as well as in the two languages, and the specific dialect features. In assessing the phonological development of a bilingual child, both languages should be the focus of attention, and each should be examined in detail, even if the child seems to be a dominant speaker of one of the languages. To this end, all phonemes of the languages should be assessed in different word positions, and phonotactic patterns should be evaluated. Assessment tools that are designed for English, no matter how perfect they are, will not be appropriate for the other language and may be the cause of over-or underdiagnosis.

Certain cases lend themselves to obvious identification of interference. For example, if we encounter in the English language productions of a Portuguese-English bilingual child forms like [t∫iz] for “tease” and [t∫ıp] for “tip,” whereby target /t/turns into [t∫], we can, with confidence, say that these renditions were due to Portuguese interference, as such substitutions are not commonly observed in developmental phonologies, and the change of /t/ to [t∫] before /i/ is a rule of Portuguese phonology.

The decision is not always so straightforward, however. For example, substitutions may reflect certain developmental simplification processes that are universally phonetically motivated and shared by many languages. If a bilingual child's speech reveals any such processes, and if the first language of the child does not have the opportunities for such processes to surface, then it would be very difficult to identify the dominant language as the culprit and label the situation as one of interference. For example, if a 6-year-old child bilingual in Spanish and English reveals processes such as final obstruent devoicing (e.g., [bæk] for “bag,” [bεt] for “bed”) and/or deletion of clusters that do not follow sonority sequencing ([tap] for “stop,” [pıt] for “spit”), we cannot claim that these changes are due to Spanish interference. Rather, these processes are among the commonly occurring developmental processes that occur in the speech of children in many languages. However, because these common simplification processes are usually suppressed in normally developing children by age 6, this particular situation suggests a delay or disorder. In this case, these processes may not have surfaced until age 6 because none of these patterns are demanded by the structure of Spanish. In other words, because Spanish has no voiced obstruents in final position and no consonant clusters that do not follow sonority sequencing, it is impossible to refer to the first language as the explanation. In such instances we must attribute these patterns to universally motivated developmental processes that have not been eliminated according to the expected timetable.

We may also encounter a third situation in which the seemingly clear distinction between interference and the developmental processes is blurred. This occurs when one or more of the developmental processes are also the patterns followed by the first (dominant) language. An example is final obstruent devoicing in the English language productions of a child with German, Russian, Polish, or Turkish as the first language. Although final obstruent devoicing is a natural process that even occurs in the early speech of monolingual English-speaking children, it is also a feature of the languages listed. Thus, the result is a natural tendency that receives extra impetus from the rule of the primary system. Other examples that could be included in the same category would be consonant cluster reduction in children whose primary language is Japanese, Turkish, or Finnish, and single obstruent coda deletion in children whose primary language is Japanese, Italian, Spanish, or Portuguese.

Besides the interference patterns and common developmental processes, speech-language pathologists must be watchful for some unusual (idiosyncratic) processes that are observed in children (Grunwell, 1987; Dodd, 1993). Processes such as unusual cluster reduction, as in [ren] for train (instead of the expected [ten]), fricative gliding, as in [wıg] for fig, frication of stops, as in [væn] for ban, and backing, as in [pæk] for pat, may occur in children with phonological disorders.

Studies that have examined the phonological patterns in normally developing bilingual children (Gildersleeve, Davis, and Stubbe, 1996) and bilingual children with a suspected speech disorder (Dodd, Holm, and Wei, 1997) indicate that children in both groups exhibit patterns different from matched, monolingual peers. Compared with their monolingual peers, normally developing bilingual children and bilingual children with phonological disorders had a lower overall intelligibility rating, made more errors overall, distorted more sounds, and produced more uncommon error patterns. As for the difference between normally developing bilingual children and bilingual children with phonological disorders, it appears that children with phonological disorders manifest more common simplification patterns, suppress such patterns over time more slowly, and are likely to have uncommon processes.

As speech-language pathologists become more adept at differentiating common and uncommon phonological patterns and interference patterns in bilingual children, they will also need to consider not only the languages of the client, but also the specific dialects of those languages. Just as there are several varieties of English spoken in different countries (e.g., British, American, Australian, South African, Canadian, Indian) and even within one country (New England variety, Southern variety, General American, and African American Vernacular in the United States), other languages also show dialectal variation. Because none of these varieties or dialects of a given language is or can be considered a disordered form of that language, the child's dialectal information is essential. Any assessment of the child's speech must be made with respect to the norm of the particular variety she or he is learning. Not accounting for dialect features may either result in the misdiagnosis of a phonological disorder or escalate the child's severity rating.

Last but definitely not least is the desperate need for information on phonological development in bilingual children and assessment procedures unique to these individuals. Language skills in bilingual persons have almost always been appraised in reference to monolingual standards (Grosjean, 1992). Accordingly, a bilingual child is assessed with two procedures, one for each language, that are designed to evaluate monolingual speakers of these languages. This assumes that a bilingual individual is two monolingual individuals in one person. However, because of the constant interaction of the two languages, each phonological system of a bilingual child may, and in most cases will, not necessarily be acquired in a way identical to that of a monolingual child (Watson, 1991).

In order to characterize bilingual phonology accurately, detailed information on both languages being acquired by the children is indispensable. However, data on the developmental patterns in two languages separately would not be adequate, as information on phonological development in bilingual children is the real key to understanding bilingual phonology. Because bilingual speakers' abilities in the two languages vary immensely from one individual to another, developing assessment tools for phonological development is a huge task, perhaps the biggest challenge for the field.

See also bilingualism and language impairment.