The present study employed exploratory factor analysis in order to describe inter-relationships between different second language (L2) production measures and to find underlying constructs that affect them. Initially, 11 L2 production measures were considered for factor analysis; however, after examining factorability, the following five L2 speech production measures were used in the analysis: (1) S-nodes per T-unit; (2) Guiraud 2000; (3) percentage of error-free T-unit; (4) hesitation episodes per 10 seconds; and (5) silent pauses per 10 seconds. Since the purpose of the factor analysis was descriptive and exploratory as opposed to descriptive and inferential (e.g., Tinsley & Tinsley, 1987), a principal factor method was used to extract underlying factors, which were subsequently rotated using a promax method. Two underlying factors were identified, the former of which was loaded by S-nodes per T-unit, hesitation episodes per 10 seconds, and silent pauses per 10 seconds, and latter of which was loaded by Guiraud 2000 and percentage of error-free T-unit. After the interpretations of factor loadings and the structure matrix, the first factor was named as Speech Incessancy and the second factor Form Consciousness. Discussion includes appropriateness of the factor names, and limitations are addressed. Studies on task based language teaching (TBLT) has attracted increasing attention to the effect of manipulating task complexity on second language (L2) learning and performance. One of the necessary consequences of this line of research is the recognition of the importance of task performance assessment. In addition to active discussions on the methods of performance assessment (e.g., Brown, Hudson, Norris, & Bonk, 2002), methods to analyze and assess L2 performance in quantitative terms have also drawn considerable attention (e.g., Ellis & Barkhuizen, 200?). For instance, general categories of L2 production measures have been discussed; they include categories of complexity, accuracy, fluency and lexis (e.g., Skehan, 2009) as well as spoken/written and frequency/ratio/formulaic measures (e.g., Wolfe-Quintero, Inagaki, & Kim, 1998). Another issue is the distinction between general and specific measures. General measures are useful when comparing effects of different types of pedagogic tasks on L2 performance whereas specific measures are sensitive to cognitive demands tasks impose (e.g., Robinson, Cardieno, and Shirai, 2009). The categories of complexity, accuracy, and fluency (CAF) are further considered as multi-dimensional (e.g., Norris & Ortega, 2009) and dynamic (Larsen-Freeman, 2009), which evade simplification. There are of course other important issues, such as reliability, validity, and sensitivity of individual L2 production measures as well as inter-relationships between L2 production measures. The present study, among these important issues, will focus on relationships between L2 production measures of complexity, accuracy, and fluency. In order to investigate relationships between different L2 production measures, factor analysis has been used. Mehnert (1988) for example, using various L2 production measures, identified three underlying factors: complexity, accuracy and fluency. Skehan and Foster (2005) identified four factors: on-line planning, complexity, accuracy and breakdown fluency. Finally, Tavakoli and Skehan (2005) found three factors: temporal aspects of fluency, repair fluency, and language form (accuracy and complexity). The goal of the present study is to explore constructs underlying various L2 speech production measures. The purpose of the present study is exploratory and descriptive, not inferential. The database used in the present study is a sub-corpus, which was originally obtained for Ishikawa (2008, in preparation).