Eye blinks: new indices for the detection of deception

Abstract

Eyeblink variables were investigated while subjects performed a guilty knowledge test (Experiment 1) and a dual modality attention task (Experiment 2). In both experiments, the temporal distribution of blinks was analyzed using an automatic video based blink analysis system [Matsuo and Fukuda, Jpn. J. Physiol. Psychol. Psychophysiol., 14 (1996), 17]. In experiment 1, the blink rate pattern discriminated between relevant and irrelevant stimuli. In experiment 2, the blink rate peak after the auditory stimulus disappeared during visually attended tasks whereas the blink rate peak after the visual stimulus was significant during auditory attended tasks. It was suggested that eye blinks could be related to the selective attention and that eye blinks could provide an additional index for the detection of deception.

Introduction

Lie detection using physiological methods has been utilized for the past 75 years with varying rates of success in identification of the guilty (Lykken, 1960). In the detection of deception the most commonly used physiological measures include respiration, electrodermal activity, cardiovascular measures and, more recently, event related potentials (ERPs). Respiratory, electrodermal and cardiac activity is principally innervated by the autonomic nervous system, and the emotional effect of deception is presumed to be reflected in such autonomic measures (Cutrow et al., 1972), ERPs which are innervated by the central nervous system are considered to be affected by the recognition of important events (Hira, 1998), a more cognitively determined activity. Oddball tasks, in which subjects must detect relatively infrequently occurring targets interspersed between non-target stimuli, are used in the ERP studies. The P300 in ERPs to target signals differ significantly from the response to non-target stimuli, and can be used to determine whether the subject is identifying target signals.

Eyeblink measures have attracted considerable attention, since they have been related to cognitive processes (Stern et al., 1984). They can be recorded easily without the subject's awareness and without the application of electrodes. Matsuo and Fukuda (1996) developed an automatic eye blink analysis system using video technology. The system allows for the identification and timing of blinks with respect to stimuli but also provides for an analysis of the eyeblink waveform. Stern et al. (1984) suggested that such measures of eyeblinks should be useful in the study of both repetitive and discrete trial tasks. In repetitive trial tasks, such as in vigilance tasks, blinks might index states of boredom and fatigue. Discrete trial tasks where the temporal relationship between blink onset and both stimulus and response onset is available allows for the analysis of phasic cognitive processes.

It was reported that blink latency, the time interval between stimulus onset and blink onset, was delayed by cognitive processes and motor responses (Goldstein et al., 1985, Bauer et al., 1985). Fukuda and Matsunaga (1983) reported that the temporal distribution of blinks (TDBs), namely the temporal change in blink rate from stimulus or response onset, varied according to tasks requiring discriminating between stimuli such as go/no go tasks. For example, blinks were inhibited just before (in anticipation) relevant stimuli, and blink bursts occurred between 0.5 and 1 s after relevant stimuli. The burst of blinks generated the blink rate peak in the TDB, with peak height increasing as a function of stimulus relevance. (Fukuda, 1994, Ohira, 1995, Ohira, 1996). Timing of the peak was delayed by information processing demands (Fukuda et al., 1991). Furthermore, the blink rate trough in the TDB occurred concurrent with key press onset, with blink rate peaking immediately following the key press (Fukuda and Matsuo, 1997). The analysis system provides information not only about traditional blink parameters such as blink latency and blink rate, but also generates new blink parameters such as blink rate peak and trough, which are variables not evaluated in the study of traditional blink parameters. The TDB resembles the ERP in averaging data across trials. However, the TDBs can be recorded without the noises and artifacts seen in conventional EEG-ERP analyses.

It was expected that the TDBs would change significantly during performance of oddball and selective attention tasks, since eyeblink activity has been demonstrated to be related to cognitive processing. The TDBs during oddball and selective attention have not been studied, although blink rates and latencies during deception have been examined (Cutrow et al., 1972). The purpose of the present study was to examine how TDBs change during performance of a detection of deception task and to investigate whether such blink measures are useful in the detection of deception.