QEEG Correlates of Interest
The popularity of television and film nationally and
internationally, along with recent interest in multimedia
technology, attests to the power of these media in attracting
people's attention and maintaining it over long periods of time.
Despite the pervasive and influential role of cinematic media
(television, films, video, multimedia) in informing, influencing,
and entertaining individuals, few studies have investigated
psychophysiological correlates of television or film viewing. A
number of concepts from various fields have been appropriated in
order to interpret psychophysiological responses to cinematic
media, but most require validation in this domain.
EEG correlates of subjective interest
Subjective interest has been studied by psychologists as
a special form of motivation (content-oriented), or a consequence
of learning (Iran-Nejad & Cecil, 1992), or a bridge between
cognition and affect (Krapp et al., 1992). Numerous
definitions of interest have been proposed. According to Reeve
(1989), interest arouses initiation and direction of attention,
followed by exploratory behavior. Interesting behaviors are
performed without anticipating external rewards and involve
enjoyment and a voluntary willingness to continue the behavior
(Reeve & Cole, 1987). Berlyne (1971) claimed that a stimulus is
interesting if it induces disorientation so that it cannot be
immediately assimilated but also contains some internal structure
so that there is a "promise of success for attempts to process
the information contained in it". In other words, interesting
information is complex but meaningful. Interest occurs when a
stimulus elicits a certain tendency to respond which is inhibited
or permanently blocked (Berlyne, 1971).
Theories of aesthetic preference
Two theories relevant to stimulus interest are proposed
by Berlyne (e.g., Berlyne, 1961; Berlyne, 1971) and Martindale
(Martindale, Moore, & West, 1988). Berlyne's psychobiological
theory maintains that stimulus preference is a function of
physiological arousal (Berlyne, 1961). Arousal depends on
stimulus properties such as intensity, meaningfulness, and
complexity. Stimulation that is complex and intense increases
arousal and is generally preferred over simplistic and weak
stimulation which curtail arousal. Excessive complexity or
intensity, however, can arouse an individual but reduce interest
or preference in a stimulus. Berlyne (1961) concluded that
stimulus preference (and interest) is related to arousal in an
inverted-"U" function (Wundt curve). Stimulus properties that
induce moderate arousal, such as modest intensity and complexity,
are most interesting and preferred.
An inverted U-shaped relationship has been demonstrated
between interest and subject characteristics. Garner and
Gillingham (1991) found an inverted "U"-shaped relationship
between interest and subject knowledge. Subjects who had moderate
knowledge about a topic were generally more interested in a topic
than those who had little or extensive topic knowledge. Others
have identified a positive linear relationship between
interestingness and sound complexity or visual pattern complexity
(Berlyne, 1971; Nicki & Gale, 1977). Berlyne (1971) concluded
that interesting stimuli may increase arousal moderately and
individuals may seek out these modest "arousal boosts".
Arousal level does not determine stimulus interest or
preference alone. Noxious stimuli may produce intermediate
arousal as readily as pleasant stimulation (Zillman, 1982).
Cognitive processes associated with arousal patterns must also be
considered. Martindale's proposed a cognitive model of stimulus
preference which claims that aesthetic preference is related to
prototypicality and meaningfulness in a (non-inverted) "U"-shaped
function. Smith and Melara (1990) found evidence of this model in
musical preference. Atypical harmonic progressions were judged
more interesting than typical progressions. Preference was also
greatest for category exemplars that are very typical or very
atypical (Martindale et al., 1988). Stimuli of moderate
typicality were least preferred by subjects.
"Arousal jag" model
McClelland (1953) argued that it is not arousal
per se that is rewarding, but fluctuation in
arousal. Individuals seek a temporary rise in arousal or tension
for the sake of pleasurable relief (i.e., drop in arousal). These
"arousal jags" were described by Berlyne (1971) as disturbance
and relief sequences. According to this model, it is predicted
that arousal variability across time will reflect subjective
interest (see Figure 5.1). Unfortunately, almost no EEG
researcher analyzes spectral parameters which can measure
temporal variability or "jags" (e.g., standard deviation,
residual variance).
Figure 5.1. Predicted relationship between alpha activity
and subjective interest according to the arousal jag model
(Berlyne, 1971).
Attentional inertia
The longer an individual maintains attention toward a
stimulus, the more likely he or she will continue to attend to
this stimulus at the expense of competing stimuli. Likewise, when
attention to a stimulus has been rather short in duration (e.g.,
task or stimulus onset), individuals are more easily distracted
by other inputs. Anderson and Lorch (1983) termed this phenomenon
"attentional inertia", which appears to be a nonstrategic
attentional phenomenon common in children and adults.
Hawkins, Yong-Ho, Pingree (1991) investigated
attentional inertia in children using normal and degraded (e.g.,
backward audio, random scenes) segments. Children maintained
attention to the first 3 s regardless of segment type. This
finding mirrors the report by Lang (1990) that onset of
television commercials elicits an orientation response in heart
rate regardless of program content. After initial processing
(e.g., 3 s), attention to a program varies as a function of
segment characteristics such as interestingness and
comprehensibility. Hawkins et al. (1991) concluded that
the initial period of attention was used by individuals to make
decisions about whether or not the program was interesting enough
to continue to view the stimulus.
According to the attentional inertia theory, all
cinematic narratives will attract attention initially. As
individuals integrate narrative content further, attention and
associated alpha activity may or may not decline depending on
subjective interest, complexity, and other properties. Two
possible predictions can be derived from this model: either the
rate of decline will reflect subjective interest or
attention may decline rapidly after initial processing to a
stable amplitude magnitude in response to narrative
properties (monotony, excitement). Figure 5.2.1 illustrates
hypothetical alpha amplitudes associated with a consistent
decline in attention during medium and low interest conditions.
This decline can be observed in trend and, to lesser extent, in
amplitude and variability. Figure 5.2.2 illustrates alpha
activity associated with rapid judgments of narrative interest
which results in a stable level of attention; subjective interest
will be reflected in amplitude.
Figure 5.2. Two possible models of the relationship
between alpha activity and subjective interest. Initial values
reflect attentional inertia.
Indexing attention
Various behavioral and subjective measurements,
including viewing diaries, recall tests, and interest ratings,
have been used to evaluate interest and attention in media
research. Unfortunately, the reliability of such measures is
questionable. It is not unusual for the variability across
subject responses to exceed the variability of underlying
psychological processes (Gescheider & Bolanowski, 1991). How
subjects assign numbers to sensations can account for more
response variance than differences in sensory processes. Social
and psychological elements can also confound subjective
responses. Subjects may inflate ratings to justify their time and
effort (Bem, 1967) or they may respond in a manner that they
believe is expected of them (Asch, 1956). Subjective measures are
less reliable and less valid than most measures because they pass
through the subject consciousness. A set of biases and competing
information have to be overcome before accurate measurements can
be obtained.
The greatest problem associated with behavioral measures is
the inability to gauge vigilance. Measures such as gaze interval
and force-choice tasks ignore the vigilant aspect of attention
(Reeves et al., 1985; Anderson, Choi, & Lorch, 1987).
Individuals can gaze at a stimulus such as a television program
in distraction, without attending to details or deeply encoding
meaningful events, or they can be scrutinizing all aspects of the
stimulus, discerning subtext and meaning. While subjective
ratings may estimate vigilance better than behavioral indices,
ratings and other subjective indices are typically crude
summaries of experiences.
Physiological correlates of story characteristics
Heart rate, systolic and diastolic blood pressure
(Zillman, 1982), skin conductance, respiratory rate, electro-
gastric responses (e.g., Baldaro et al., 1990) have all
been used to index attentional or processing correlates.
Vigilance levels can be evaluated by the increases or decreases
in arousal and for short intervals of time. Heart rate has been
shown to respond to points of interest in a film such as close-
ups and movements (Lang, 1990). Electrogastrographic rates and
heart rate both increase in response to emotional scenes in films
(Baldaro, Battacchi, Trombini, Palomba, & Stegagno, 1990). The
P300 amplitude at the mid-parietal area can differentiate
exciting moments in films from boring moments, albeit with
multiple replications (Rosenfeld, Bhat, Miltenberger, & Johnson,
1992).
EEG correlates of film viewing
Although continuous EEG has not yet been used to assess
subjective interest, emotional and attentional responses to films
have been associated with continuous patterns of EEG. Greater
nonspecific activation is observed when negative or positive
films (e.g., sick children, children playing) are watched
compared to neutral films (Schellberg, Besthorn, Thomas, &
Gasser, 1990; Reeves et al., 1989). Negative television
scenes activate right frontal cortex and positive scenes activate
left frontal cortex (Reeves et al., 1989; Jones & Fox,
1992). Individual differences in anterior EEG asymmetries are
predictive of affective responses to cinematic stimuli (Tomarken
et al., 1990; Jones & Fox, 1992). In addition, how well an
individual encodes a television program can be estimated by
measures of topographic EEG (Reeves et al., 1985).
Advantage of topographic EEG
Any psychological phenomena that fluctuates across time
will be better assessed by EEG recordings, which possess temporal
resolutions below 1 s, than by behavioral or subjective, measures
which typically involve intervals from 10 s to 60 min or more.
EEG measures also possess functional resolution. Most
neuropsychological tests such as dichotomous listening assess one
or two functional processes or regions. Topographic EEG denotes
functional activity from multiple regions of the brain
simultaneously. EEG also possesses numerous dimensions of
information, such as amplitude, frequency, and spatial coherence,
giving it potentially greater discriminatory power.
Although topographic EEG may be more expensive to
administer per subject; for the additional information obtained,
the cost is comparable if not superior to subjective and
behavioral techniques. In this study, quantitative topographic
EEG possesses two significant advantages over subjective and
behavioral measures: passivity and functional resolution. Not
only can processing changes be localized to specific cortical
areas, because subjects are not conscious of their
psychophysiological responses, identification of high and low
interest states can be achieved unobtrusively, without altering
film-viewing habits.
Summary of predictions
Temporary increases in attention at the onset of new
stimuli may be an innate response (orienting reflex) and should
not reflect conscious interest in a stimulus (Hawkins et
al., 1991; Reeves et al., 1985). Extended attention to
a film, however, will reflect interest (Anderson & Lorch, 1983;
Berlyne, 1961). According to the arousal jag model, high interest
films will be associated with elevated variability and trend
instability (see Figure 5.1). Low and medium interest films will
result in reduced variability, higher alpha amplitudes and
positive trends (see Figure 5.2). High interest films will result
in nonspecific activation as well as functional engagements due
to auditory (temporal), visual (occipital), and multimodal
integration (parietal lobe).
METHOD
Subjects
EEG was analyzed from the same 20 subjects described in
Chapter 3. An additional group of 32 subjects (18 males and 14
females) between 21 and 40 years of age (mean age of 20.4 years)
were shown all 27 film trailers to judge independently subjective
interest elicited by each film.
Materials and Apparatus
Questionnaires and film previews were those used in
Chapter 3. An additional questionnaire was used to estimate actor
familiarity. Subjects rated actors and actresses who starred in
each film on familiarity using a 7-point Likert-type scale with
the following scale anchors: 0 not at all familiar to 6
very familiar. The highest rating for each trailer (of the
two or three actors present) determined actor familiarity rating
for the trailer. EEG acquisition, preparation, and apparatus were
identical to those presented in Chapter 3.
Procedure
The procedure presented in Chapter 4 was used. Of the 15
films viewed, approximately five trailers were from each interest
level group (high, medium, and low). After completion of the
study, half of the subjects (5 male, 5 female) rated familiarity
of actors starring in all films. This procedure confirmed that
interest in a trailer was not related to familiarity of actors
who starred in the film [r= -.06, F(1,23)= 0.091, ns]. Data
analysis followed principles established in earlier chapters
RESULTS
Three levels of subjective interest (HIGH, MED, LOW) and
replicated eyes open baseline (EO) were compared in four spectral
parameters of alpha activity (8-12 Hz). Condition means consisted
of 51 to 71 epochs and were compared using univariate ANOVAs with
no correction for independent tests. Data were initially pooled
into functional areas and degrees of freedom were adjusted.
Nonspecific differences between EO and interest
conditions
As shown in Table 5.1, main effects were seen in three
spectral parameters [FLMAGN(3,57)= 4.842, p<.05;
FSD(3,57)= 9.573, p<.05; FSLOPE(3,57)= 3.538,
p<.05; FRV(3,57)= 1.396, ns]. EO differed from all
subjective interest levels in variability, from MED and HIGH in
amplitude, and from HIGH only in trend [p<.05].
Table 5.1. Mean spectral values of subjective interest (19 sites, n=20)
LMAGN SD SLOPE RV
--------------------------------------------------------
EO 1.41 1.92 0.010 0.013
LOW 1.37 1.70* 0.003 0.013
MED 1.35* 1.60* 0.004 0.013
HIGH 1.32* 1.54* -0.000* 0.012
--------------------------------------------------------
* p<.05, compared to EO
Specific differences between EO and interest conditions
Condition by recording site interactions were observed
in variability and trend instability [FLMAGN(4,67)= 1.612,
ns; FSD(2,33)= 4.298, p<.05; FSLOPE(2,33)=
1.718, ns; FRV(2,39)= 3.518, p<.05]. EO was more
variable at site T6 compared to LOW interest films
[p<.01]. EO differed from MED at sites F7, Fz, F8, P4, T5,
T6, O1, and O2 [p<.01]. EO differed from HIGH at all sites
except Fp1, Fp2, F7, C3, and T4. In trend instability EO
differed from MED and HIGH at site T6 only [p<.01].
Laterality differences between EO and interest
conditions
In terms of laterality, a main effect was found in SD
[FSD(3,57)= 3.308, p<.05] and site interactions in
SD and RV were found[FSD(7,125)= 2.169, p<.05;
FRV(8,147)= 2.110, p<.05, respectively]. Subjects
were relatively less engaged in the right hemisphere for EO and
more engaged for high interest conditions. A similar effect was
found in trend instability.
Correlates of subjective interest
Mean spectral values are reported in Tables 5.2.1 and
5.2.2 for each recording site. Main effects of subjective
interest were found in amplitude and
variability [FLMAGN(2,38)= 15.662, p<.05;
FSD(2,38)= 8.633, p<.05; FSLOPE(2,38)= 2.431,
ns; FRV(2,38)= 1.587, ns]. HIGH was more activated than
MED, which was more activated than LOW [p<.05]. HIGH and
MED were less variable than LOW [p<.05], but did not
differ from each other [p>.05].
Table 5.2.1. Topographic spectral values of subjective interest (n=20)
Log Magnitude Standard Deviation
--------------------------------------------------------
LOW MED HIGH LOW MED HIGH
--------------------------------------------------------
Fp1 1.36 1.32* 1.31 1.79 1.68 1.65
Fp2 1.35 1.27 1.30 1.78 1.63 1.63
F7 1.15 1.10* 1.10 1.26 1.13 1.15
F3 1.38 1.37 1.34 1.52 1.44 1.37
Fz 1.44 1.43 1.41 1.60 1.53 1.47
F4 1.37 1.36 1.33 1.52 1.51 1.37
F8 1.11 1.08 1.06 1.22 1.16 1.13
T3 1.13 1.11 1.09 1.29 1.22 1.18
C3 1.46 1.45 1.42 1.85 1.77 1.72
Cz 1.52 1.51 1.48 1.88 1.78 1.68
C4 1.41 1.40 1.37 1.83 1.74 1.67
T4 1.13 1.15 1.12 1.32 1.31 1.25
T5 1.34 1.31 1.28 1.69 1.56 1.51
P3 1.54 1.51 1.47* 2.14 2.02 1.94
Pz 1.57 1.54* 1.49* 2.23 2.11 1.95
P4 1.49 1.47* 1.42* 2.07 1.94 1.82
T6 1.31 1.30 1.26* 1.62 1.52 1.46
O1 1.51 1.48 1.45 1.86 1.72 1.70
O2 1.48 1.45* 1.43 1.85 1.68 1.66
--------------------------------------------------------
* p<.01, compared to condition to the left
Table 5.2.2. Topographic spectral values of subjective interest (n=20).
Slope Coefficient Residual Variance
--------------------------------------------------------
LOW MED HIGH LOW MED HIGH
--------------------------------------------------------
Fp1 0.003 0.001 0.002 0.014 0.013 0.014
Fp2 0.003 0.001 0.001 0.014 0.013 0.013
F7 0.001 0.002 0.001 0.010 0.009 0.009
F3 0.005 0.003 0.000 0.012 0.011 0.011
Fz 0.005 0.003 -0.000 0.012 0.012 0.012
F4 0.004 0.002 0.000 0.012 0.012 0.011
F8 0.000 0.001 -0.001 0.009 0.009 0.009
T3 -0.001 0.003 0.001 0.010 0.010 0.010
C3 0.003 0.007 -0.000 0.014 0.014 0.014
Cz 0.003 0.005 0.000 0.014 0.014 0.014
C4 0.003 0.005 -0.002 0.014 0.014 0.013
T4 -0.001 0.001 0.001 0.010 0.010 0.010
T5 0.003 0.005 -0.001 0.013 0.012 0.012
P3 0.003 0.007 -0.001 0.016 0.016 0.016
Pz 0.003 0.008 -0.001 0.017 0.016 0.016
P4 0.003 0.007 -0.002 0.016 0.015 0.015
T6 0.001 0.006 0.001 0.013 0.012 0.012
O1 0.003 0.007 -0.002* 0.014 0.014 0.014
O2 0.003 0.006 -0.001 0.014 0.013 0.013
--------------------------------------------------------
* p<.01, compared to condition to the left
Topographic interactions were observed in amplitude and
trend [FLMAGN(3,48)= 2.702, p=.056; FSD(2,36)=
0.735, ns; FSLOPE(3,57)= 3.323, p<.05;
FRV(2,39)= 0.880, ns]. As shown in Figure 5.3, amplitude
decreased as interest increased at sites Pz and P4
[p<.01]. HIGH films result in lower LMAGN compared to MED
films at sites P3 and T6 [p<.01] and compared to LOW films
at every site except T3 and T4. For MED films, LMAGN values were
lower at sites Fp1, F7, and O2 compared to LOW [p<.01].
Figure 5.3. Topographic differences in amplitude as a
function of subjective interest (* p<.01).
Differences in trend were less clear. Subjects
deactivated to a greater extent at site O1 in MED compared to
HIGH [p<.01] although the LOW slope coefficient was not
below the MED value but intermediate.
No laterality effects were found for subjective interest
[p>.05].
DISCUSSION
Nonspecific arousal was generated by interest conditions
compared to eyes open baseline. The right posterior temporal
cortex was engaged by even low interest films. Right posterior
cortex was activated more than analogous areas in the left
hemisphere.
EEG correlates of narrative integration
When individuals watch television, they perceive the
gist or general meaning of a scenes before putting details into
place (Graber, 1985). Holistic processing precedes analytical
processing. This may explain why right hemisphere mechanisms were
engaged by even low interest films. Cortical functions in the
right hemisphere play an important role in processing configural
or contextual information (Bogen & Bogen, 1983; Mazziotta,
Phelps, Carson, & Kuhl, 1982; Rehak et al., 1992). Alpha
attenuation at site T6 may also reflect occasional excitement
experienced during even low interest films (Lorig & Schwartz,
1989) as well as visual processing (Jones-Gotman & Milner, 1978;
Grillon & Buchsbaum, 1986).
Moderately interesting films activated mid- and right-
parietal, right occipital, and left fronto-temporal cortex. In
addition to these brain areas, high interest films also activated
the left parietal cortex. These findings support the "arousal
boost" hypothesis proposed by Berlyne (1971). Right occipital
activation associated with heightened interest typifies increased
visual processing (Van Winsum, Sergeant, & Geuze, 1984). Findings
in standard deviation and residual variance did not support
Berlyne's model of arousal jags, however, but did support the
attentional inertia theory. Variability increases due to interest
may have been diminished (i.e., nullified) by the reduction in
variability associated with activation. Slope coefficients
associated with subjective interest, however, did not support the
attentional inertia model presented in Figure 5.2.1, but were
consistent with the model in Figure 5.2.2.
Fronto-temporal activation
Left fronto-temporal cortex plays a crucial role in
processing temporal information. Patients with left fronto-
temporal lesions exhibit deficits in self-ordered tasks (Petrides
& Milner, 1982), temporal rule-induction tests (Villa et
al., 1990), and other tasks in temporal information must be
monitored (e.g., Milner & Petrides, 1984). Activation of left
fronto-temporal cortex may reflect systematic integration of
fictive events into chronologies in moderately interesting
narratives whereas this process occurred sporadically in low
interest narratives. High interest films did not result in
additional activation in this brain region, suggesting that other
content unrelated to temporal relations generated increased
interest. Disengagement in the left fronto-temporal cortex may
also indicate that subjects were very relaxed during the viewing
of the low interest films (Lorig & Schwartz, 1989).
Parietal activation
Mid- and right-parietal cortex were also activated
during moderately interesting films. This could reflect general
integrative or attentional differences between conditions,
engagement of specific mechanisms required for narrative
comprehension, or both.
Although patients with parietal lesions have difficulty
attending to stimuli regardless of side of lesion (Ladavas, Del
Pesce, & Provinciali, 1989), right-sided lesions often produce a
hemi-neglect syndrome as the left side of space is no longer well
represented (de Renzi, Gentilini, & Barbieri, 1989; Weintraub &
Mesulam, 1988). For instance, patients with right parietal
lesions omit the left side of drawings, misjudge distances of
objects on the left site, neglect sounds from the left, have
difficulty visualizing the left-side in imagery, etc. (see Stein,
1992, for review). In general, right parietal patients have
difficulty maintaining a high level of alertness as they are
unable to regulate attention (Ladavas et al., 1989). Right
parietal engagement during the films may reflect attentional
deployment to interesting stimuli.
Specific cognitive processes may also be elicited by
subjective interest. Right hemisphere patients have difficulty
judging interest level of a story, though their capacity to
encode and recall details is normal (Rehak et al., 1992).
Mead and McLaughlin (1992) investigated eye fixation and
discovered that preference to pictures occurs more often when the
majority of a picture is situated in the left visual field,
engaging right hemisphere functions. Interest is consistently
affiliated with two properties, novelty and complexity (Berlyne,
1970; Berlyne, 1971; Garner, 1992). Right hemisphere mechanisms
are critical in processing complex configural information (Bogen
& Bogen, 1983) as well as novel stimulation (Bradshaw &
Nettleton, 1981; Goldberg & Costa, 1981).
Finally, high interest films resulted in left parietal
activation, reflecting greater attention to verbal content or an
analytical approach in the interpretation of cinematic narratives
(Bradshaw & Nettleton, 1981; Dujardin et al., 1993).
Mean amplitude values reflect gross changes associated
with stimulus interest. In Chapter 6, other characteristics of
the EEG signal, such as the pattern of modulation, will be
investigated.
|
