Elicited Imitation in Second Language Acquisition Research

This review surveys theoretical and empirical evidence about elicited imitation (EI) as a measure of second language (L2) learners' performance. Different from natural imitation, where children repeat utterances without request in a natural setting, EI requires participants to hear and then repeat a sentence that usually includes a target grammatical structure in a laboratory setting. EI is assumed to be reconstructive; that is, during EI, participants go through cognitive processes such as (a) processing a stimulus sentence, (b) reconstructing it with their own grammar, and (c) reproducing it. Additional evidence that EI is reconstructive and not rote repetition is that speakers who have internalized the target structures spontaneously correct ungrammatical EI stimuli (e.g., Erlam, 2006; Hamayan, Saegert, & Larudee, 1977; Munnich, Flynn, & Martohardjono, 1994).

Since the 1960s, EI has been widely used in (a) first language acquisition research (e.g., Fraser, Bellugi, & Brown, 1963; see Lust, Flynn, & Foley, 1996; Slobin & Welsh, 1973 for reviews), (b) assessment of first language disorders in children (e.g., Dailey & Boxx, 1979), and (c) second language acquisition (SLA) research (e.g., Naiman, 1974). This review is limited to SLA studies.

EI has been used to assess L2 competence¹ (see Erlam, 2006; Vinther, 2002 for reviews) since Naiman's (1974) seminal work. SLA studies have focused primarily on the assessment of English performance (e.g., Ellis, 2005, 2006; Ellis, Loewen, & Erlam, 2006; Erlam, 2006; Gallimore & Tharp, 1981; Hamayan, Saegert, & Larudee, 1977; Henning, 1983; Munnich et al., 1994; Perkins, Brutten, & Angelis, 1987; Spitze & Fischer, 1981), although it has been used, albeit infrequently, to assess other languages such as French (e.g., Burger & Chrétien, 2001; Markman, Spilka, & Tucker, 1975), Spanish (e.g., Scott, 1994), and Dutch (e.g., Van Boxtel, Bongaerts, & Coppen, 2005). As such, EI has been used to collect data that reflect L2 learners' performance in SLA.

Advantages

EI provides a number of advantages. First, a wide range of L2 structures can be elicited relatively easily. While most researchers have tapped into syntax (e.g., Ellis et al., 2006; Erlam, 2006), pronunciation (e.g., Burger & Chrétien, 2001; Henning, 1983) and discourse markers (e.g., Burger & Chrétien) have also been elicited. Second, researchers have a degree of control over administration and analysis with the flexible and reliable procedures of EI.² This allows SLA researchers to elicit target forms orally, although that has been difficult to accomplish in the past³ (Ellis & Barkhuizen, 2005; Mackey & Gass, 2005). Third, EI can be used with different age groups (e.g., children and adults), different languages (e.g., English, French, and Dutch), and different populations (e.g., Chinese, Japanese, Hawaiians, and Canadians). Finally, when compared with other tasks, EI has been shown to have high predictive and Rasch Fit validities (Henning, 1983).

Challenges

In the 1970s, EI's validity was challenged: the major criticism being the possibility of rote repetition in response to stimuli (i.e., participants may be simply parroting what they hear). For EI to examine L2 performance rather than short-term memory capacity, most researchers carefully control the length of stimulus sentences by manipulating the number of syllables or words in the sentences (see Vinther, 2002, for a review). Although researchers agree that they must ensure that stimuli exceed short-term memory capacity, there is no consensus on the appropriate length of the variables that accurately represents participants' performance.

The effect of short-term memory capacity is also of great concern in the methodology of EI (Bley-Vroman & Chaudron, 1994). Regardless of the duration of a stimulus, the first and last items are recalled better than the middle items (i.e., serial order effect). Additionally, researchers must be careful to avoid floor and ceiling effects of EI tasks because the stimuli for EI may be difficult or easy. If accuracy performance during EI is too high or too low, researchers are not certain about what they are actually measuring.

Linguistic complexity affects the results of EI as well. As Smith (1973) and Hamayan et al. (1977) claim, some grammatical structures may be easy to repeat (e.g., conjunctions, complements, and number constructions...