R. PFISTER 267
-50
0
50
100
150
200
250
300
350
400
Search Volume Index
Wee k I ndex ( 2004)
February 2004:J anet Jac ks on' s
wardrobe malfunction during live
performance at the Super Bowl
show on January 31.
Figure 1.
Search Volume Indices (SVI) as a non-reactive, relative measure of
internet traffic produced by Google searches for “Janet Jackson” in the
course of 2004. SVIs can be interpreted as a week’s traffic in relation
to the average traffic during January 2004. The first week of February
2004—following Jackson’s wardrobe malfunction during the Super
Bowl show on January 31—thus yielded an increase of relative internet
traffic by a factor of about 370.
month (fixed scaling). Thus, a fictive SVI score for the term
“Psychology” of 1.5 in the last week of August 2004 indicates
that internet traffic produced by this term in this week is 1.5
times as much as during January 2004. Fixed SVIs were ex-
tracted for each of the four weeks prior to the event (being rela-
tively constant at SVI = 1), for the week of the event, and for
the following eight weeks.
Results
All scores were baseline-corrected by subtracting the mean
SVI of the four weeks prior to the wardrobe malfunction. SVIs
were averaged across all 17 cases and 99% confidence intervals
were computed for each week (see Figure 2). SVIs instantly
increased in the week of the event and stayed at a constant level
for the next week, too. In week three after the event, SVIs were
markedly decreased but still significantly above baseline,
marking the end of the wardrobe plateau. Finally, and being at a
relatively low level already, SVIs started to decrease asymp-
totically in the following weeks (see Figure 2; details of the
effects for each case of wardrobe malfunction are included in
the Appendix).
For further statistical analysis, individual profiles were ag-
gregated to four time markers: the time prior to the wardrobe
malfunction (baseline), the week of the wardrobe malfunction,
the week after the wardrobe malfunction, and the following 7
weeks. These four data points were analyzed by within-subjects
ANOVA with the single factor of time. The factor’s significant
influence, F(3, 48) = 3.73, p = .035, ηp2 = .19 (Greenhouse-
Geisser adjusted to account for non-sphericity), was mainly
based on a quadratic trend with timescale-corrected contrast
vector, F(1, 16) = 7.16, p = .016, ηp2 = .31, without linear or
cubic components (both F’s < 1).
Discussion
The impact of wardrobe malfunctions on a celebrity’s inter-
net presence was assessed by means of the internet traffic that
was produced by Google searches for the name of the celebrity
in question. Examining 17 cases of wardrobe malfunction, a
statistically reliable increase of searches was found as long as
the third week after the incident with a plateau activity during
-1
0
1
2
3
4
5
‐4‐3‐2‐1012345678
Search V ol ume Index
Wee k Index
Figure 2.
Baseline-corrected mean SVI scores of 17 cases of wardrobe malfunc-
tion in relation to the onset of the event in week 0. Error bars represent
99% confidence intervals.
the first two to three weeks. This plateau activity amounted to a
baseline-corrected SVI of about 3, indicating a substantially
increased internet presence during that time. Given the
fast-paced nature of the internet community, the reported im-
pact duration of about 3 weeks is even more remarkable and
indicates a profound interest in this type of news among a high
proportion of internet users. Even though no conclusions are
possible about the interrelations with other factors, the motive
of using the internet as a source for celebrity gossip in combi-
nation with soft pornographic material fits with other results on
internet-related motives such as social and affective motives or
interest in pornography (Young et al., 1999).
Considering the nature of SVIs as dependent variable, sev-
eral alternative explanations might also account for the in-
creased interest in a celebrity after a wardrobe malfunction. For
instance, the present experiment only analyzed events that were
broadcasted to numerous people. Thus, the increased interest
might indicate solely an interest in the event as such. This al-
ternative explanation was tested by further analyses that used a
similar situation but without wardrobe malfunction and applied
the same test as before (data not shown here). In these analyses,
no comparable increase could be detected for any event. The
increased internet presence of a celebrity after a wardrobe mal-
function is thus indeed specific to the event studied.
Taken together, the present study demonstrated the use of
Search Volume Indices as a non-reactive measure of internet
behaviour. The measure provides an attractive opportunity for
internet research as the data is readily available and provides a
unique opportunity for studying internet behaviour on a broader
scale. The method can easily be applied to a wide variety of
research questions and might provide a valuable tool for media
researchers.
References
American Dialect Society (2005). 2004 Words of the year final vote.
URL (Retrieved 5 November 2009) http://www.americandialect.org
/2004_Words_of_the_Year_Final_Vote.pdf
Choi, C. J., & Berger, R. (2010). Ethics of celebrities and their increas-
ing influence in 21st century society. Journal of Business Ethics, 91,
313-318. doi:10.1007/s10551-009-0090-4
Gamson, J. (2000). The web of celebrity. American Prospect, 11,
40-41.
Google Inc. (2009). About Google Trends. URL (Retrieved 25 October
2009) http://www.google.com/intl/en/trends/about.html#7
Holland, S. L. (2009). The “offending” breast of Janet Jackson: Public
discourse surrounding the Jackson/Timberlake performance at Super