Contemporary researchers have suggested that our propensity to engage in off-task thoughts and to suffer performance failures (i.e., to experience lapses) is strongly related to our ability to engage attentional control (AC). As a construct, AC has been defined as the ability to maintain attention on task goals and to sustain attention in the face of internal and external distractions. However, the construct validity of AC remains untested, with earlier research suggesting it is comprised of variance from both goal maintenance and alertness tasks. Furthermore, recent evidence has suggested that tonic and phasic alertness, as measured by baseline and task-evoked pupil diameters, are predictive of lapses. The current study was thus motivated by the need to further understand the attentional mechanisms underlying lapses. Further, because AC partially explains the relationships between working memory (WM) and fluid reasoning (gF), it was important to explore how the attentional mechanisms underlying lapses related to WM and gF.Three measures (or manifest variables) for each of the six main cognitive constructs: tonic alertness, phasic alertness, goal maintenance, lapses, WM, and gF were collected in a heterogeneous community sample of adolescents, aged 13 to 18 years. Exploratory factor analyses were used to determine the construct validity of tonic alertness, phasic alertness, and lapse constructs. Confirmatory factor analyses and structural equation modeling were used asses the relationship between all six cognitive constructs. Results indicated that retrospective questionnaires, thought probes, and performance-based measures of lapses converged onto six distinct factors. However, only performance-based lapses were significantly related phasic pupil-based alertness, AC, WM, and gF constructs. Pupil- and performance-based measures of alertness did not converge, indicating that these alertness measures may not measure the same construct. Additionally, tonic and phasic alertness emerged as separate constructs. The construct validity of AC was verified as a combination of tonic alertness and goal maintenance abilities. Structural equation modeling revealed that phasic alertness (i.e., mean task evoked pupil diameters) predicted lapses (i.e., reaction time variability and errors) during attention demanding tasks. Replicating prior research, lapses predicted AC, AC predicted WM, and WM predicted gF. Theoretical implications of these results will be discussed.