Complexity, Accuracy, and Fluency in L2 Speaking

This section addresses spoken development among the learners in the Pitt IEP through the lens of how learners improve in their (global) spoken proficiency in an activity which the IEP calls a ‘recorded speaking activity’. As part of the Pittsburgh Science of Learning project (www.learnlab.org), the IEP encouraged research on spoken production, including complexity, accuracy, and fluency (see an overview of this research) and the importance of noticing and feedback (Long, 2007; Schmidt, 1990). But before delving into the CAF research, it is useful to briefly review just how complex spoken production is and how instruction in the CAF framework targets different facets of the speech production process.

Speaking is a complex skill for expert users and even more challenging for L2 speakers because it involves the coordination of many sources of knowledge in real time. In any language, speech production involves (a) conceptual planning, (b) selection of vocabulary and syntax to represent those concepts, and then (c) articulatory production that converts a nonlinear order into a linear sequence, and (d) monitoring. Stages (b) and (c) can overlap, which contributes fluency in speaking as grammatical-lexical selection is occurring at the same time as articulation (Kormos, 2006, pp. xviii-xix; Levelt, 1989). Unlike writers, who can carefully revise a sentence, speakers have only limited time to check the output that they want to produce. This time pressure often results in speech errors that reveal the underlying production processes (e.g., papers in Fromkin, 1973). For example, spoonerisms, in which a sound in a word ‘moves’ to another word (either earlier or later) (e.g., I have a well-boiled icicle from I have a well-oiled bicycle) tell us that speech planning affects even the linear order of single sounds as well as words. MacKay (1973, p. 192) suggests that such speech errors imply that lexical items are put into a ‘buffer’ and broken down into the phoneme level before motor output (movingthe mouth in real-time pronunciation). These errors explain in part how it is that phonemes (and words themselves) can be re-ordered in sentences during speech production. If speaking one’s first language is such a complex process, no wonder second language learners’ production can be slow and often lags behind their comprehension. The specific details of the psycholinguistics of (bilingual) speech production have been discussed in depth in Kormos (2006), and readers should be aware of the rich literature on this topic if they wish to pursue more sophisticated pedagogical interventions and understand all the processes that can influence bilingual speech production.

In a way, the CAF approach to L2 spoken proficiency concentrates on stages (b) and (c) of speech production. Although a great deal has been written about CAF, especially in the context of task-based language instruction (TBLT; see Skehan (2018) for a recent overview), the definition of each component and the best way to measure each one is not always agreed on. First, accuracy is fairly obvious as in the absence of errors, but, as we saw in the previous chapter, the grain size of the grammatical structures being analyzed and the counting of ‘errors’ can be a problem in accurately tracking LI influence in development of accuracy. Overall, and perhaps surprisingly, it seems that ‘error’ is a term that is reserved for an error in morpho-syntax and that pronunciation errors are not considered separately. Also, the possibility that an error with phonological structure could influence morpho-syntax is also not often considered. One exception is Goad (2011), who analyses some morphological errors in L2 English as arising from constraints on phonological word formation. Moreover, the distinction between ‘error’ and ‘mistake’ is also not dealt with in depth in measuring accuracy (Towell, 2012, p. 52). Thus, even a measure as apparently straightforward as ‘accuracy’ turns out to be a challenge.

Fluency is operationalized in various ways, for example, as a function of the rate of speech (e.g., amount of speech within a set period); the number of pauses, both filled (urns and ers in English) and unfilled (when the speaker simply stops and silence ensues); these pauses can occur in the middle of a clause or between speech units; and, finally, the number of false starts and repetitions.

Complexity is perhaps the most challenging construct, as it may be hard to establish how complexity is determined (Bulte & Housen, 2012). The influence of the tasks on the measure of complexity can be important (Robinson, 1995, 2011a, 2011b). However, questions remain on whether complexity can increase naturally as proficiency increases or whether complexity is dependent on tasks that learners are asked to carry out. Despite such concerns, complexity is generally held to be a measure of the learner’s interlanguage grammar; that is, the range of linguistic structures available to learners in their grammars (e.g., control of tense and aspect; noun clauses, adjective clauses; quantifiers) and how the learner demonstrates this knowledge in real-time spoken performance to achieve certain functional goals (e.g., describing, persuading, narrating) (Housen & Kuiken, 2009, p. 464).

6.2.1 Recorded Speaking Activities in the Pitt Intensive