Measured Language: Quantitative Studies of Acquisition, Assessment, and Variation

8. Speedup versus Automatization: What Role Does Learner Proficiency Play?

1 1 1 8 ■Speedup versus Automatization What Role Does Learner Proficiency Play? JeSSicA G. cox And Anne M. cAlderón Georgetown University ■ The TransiTion from conTrolled processing to automatized processing in L2 acquisition is characterized by moving from slow, effortful processing to quicker processing outside of conscious control. However, this has been difficult to operationalize since it is not reportable. The current work builds on Hulstijn, van Gelderen, and Schoonen’s (2009) attempt to operationalize and quantify the two. The distinction between automatization and speedup is important within the SLA field because it reflects a change in the learner’s mind. Phillips et al. (2004) reported that the more proficient the learner, the higher the level of automatization; however, Hulstijn, van Gelderen, and Schoonen (2009) failed to find the same result in a longitudinal study. Previous studies, however, have not included independent assessments of L2 proficiency nor have they controlled for lexical knowledge. The present study investigated the role played by L2 proficiency in the transition from speedup of controlled processes to automatization. One hundred and three L1 English university students enrolled in either intermediate, advanced, or advanced+ Spanish, completed L1 and L2 semantic decision tasks that recorded speed and accuracy (based on Segalowitz and Freed 2004). Proficiency groups were determined by class level and justified with an independent measure (Diploma de Español como Lengua Extranjera) taken by a subset of participants. Speedup and automatization were operationalized following Hulstijn, van Gelderen, and Schoonen (2009). Results showed evidence of both speedup and automatization at all proficiency levels ; Fisher’s z analyses revealed that both automatization and speedup decreased in the advanced+ group. We conclude that there are significant amounts of automatization and speedup at all levels, though we hint at a change between advanced and advanced+ proficiencies. Theoretical Foundation The acquisition of a first (L1) and second language (L2), although different in many ways, are both examples of the acquisition of skills occurring in a gradual manner. 112 Jessica G. Cox and Anne M. Calderón Automatization, or the routinization and restructuring of component processes, is a key element of skilled behavior. There are two main theories that attempt to define automaticity as it relates to skill acquisition in psychology literature. Both consider automatic processing to be effortless. Anderson’s (1983) adaptive control of thought theory posits that automatization begins with controlled and conscious processes of declarative knowledge and slowly proceeds to processing of routines without attention. In contrast, Logan (1998) views automatization as starting off with rules of thumb that eventually progress to higher order “instances.” These instances finally become strong enough to cause a bypass of rule application and the learner depends solely on the retrieval of the stored instances. Fast processing, as coined by Segalowitz and Segalowitz (1993), is a concept central to skill acquisition. While automatization involves the “bypassing of serial execution of component processes” (Hulstijn, van Gelderen, and Schoonen 2009, 557), fast processing (speedup) is the “speeding up of essentially all component processes that make up the execution of a task in the earliest stage of skill acquisition” (Hulstijn, van Gelderen, and Schoonen 2009, 557). Measuring automaticity and speedup requires an understanding of the measurement of skill acquisition, which can be demonstrated empirically by a decrease in time required to perform an acquired action as practice in the skill increases (Hulstijn, van Gelderen, and Schoonen 2009). Accuracy and speed (reaction time [RT]) on speeded linguistic tasks such as lexical and semantic decision tasks are used to measure linguistic skill. The relationship between mean RT and the standard deviation of the mean (SDRT ) is, in most cases, linear: with practice, participants reduce their RT and reduce the variability in their RT (i.e., SDRT ) (Hulstijn, van Gelderen, and Schoonen 2009). Given this, Segalowitz and Segalowitz (1993) distinguished between automatization and speedup by proposing that in the case of speedup, mean RT and SDRT will each be reduced without a change in the coefficient of variance (CVRT ), calculated as SD divided by mean RT. CVRT is a measure of the variability at each level of latency (RT); in other words, it measures the efficiency of processing. In contrast, since automatization is characterized by the routinization or elimination of component processes (Hulstijn, van Gelderen, and Schoonen 2009), in this case processing is more efficient, so mean RT, SDRT , and CVRT will all be reduced. This also yields an increasing correlation between mean RT and CVRT . Research investigating speedup and automatization can be cross-sectional (individuals at differing...