Mobile phones are good for you, p 

Preview:

Citation preview

A

hpoor©

K

staimg

tdduoo

tSltc

r

0d

Neuropsychologia 45 (2007) 1580–1581

Note

Mobile phones are good for you, p < 0.36! Observationson Keetley, Wood, Spong and Stough (2006)

Michael B. LewisSchool of Psychology, Cardiff University, Park Place, Cardiff CF10 3AT, UK

Received 24 July 2006

bstract

Keetley et al. [Keetley, V., Wood, A. W., Spong, J., & Stough, C. (2006). Neuropsychological sequelae of digital mobile phone exposure inumans. Neuropsychologia, 44, 1843–1848] report a study into the effects of mobile phones on a variety of cognitive tasks. They report that theresence of an active mobile phone reduces performance on some tasks but, surprisingly, improved performance on others. Bonferroni correction

f this multi-hypothesis exploratory research, however, reveals that none of the findings reach normally accepted levels of significance. The resultsf the study remain interesting in suggesting hypotheses for further research, however, care must be taken if making conclusions based on thisesearch.

2006 Elsevier Ltd. All rights reserved.

0iwn

cmMsTnadtnfwre

eywords: Cognition; Electromagnetic field; Reaction times; Working memory

Keetley, Wood, Spong, and Stough (2006) report a majortudy into the effects of mobile phone on a range of cogni-ive tasks. Their results, like much of the literature in the field,re mixed. Some tasks show an impairment, others show anmprovement. These findings will be of particular interest to

obile-phone companies and users and the article has alreadyenerated media interest.

While the research is worthwhile and the results are useful,he strength of the conclusions made is questionable given theata reported. The claim that the study ‘provides statistical evi-ence of a cognitive difference in performance’ (p. 1847) whensing mobile phones is premature given the statistical limitationsf the study. These limitations come from the exploratory naturef the research combined with a lack of a Bonferroni correction.

The research reports 18 tests, each of which was hypothesisedo be affected by the proximity of an activite mobile phone.even of the hypotheses are supported by the data at an alpha

evel of 0.05, five showing an impairment in performance andwo showing an improvement when a mobile phone was active

lose to the participant.

If a standard Bonferroni correction were applied to thisesearch then the critical p-value would drop from 0.05 to

E-mail address: Lewismb@cf.ac.uk.

tfs

rc

028-3932/$ – see front matter © 2006 Elsevier Ltd. All rights reserved.oi:10.1016/j.neuropsychologia.2006.12.001

.002778 (=0.05/18). The smallest p-value reached in the datas 0.004 (i.e. not significant) and many are larger than this. Thisould indicate that none of the seven effects observed reach theormally accepted level of significance.

More specifically, let us consider the particularly strikinglaim made in the abstract, which is bound to be picked up by theedia, that mobile phones can improve performance in a Trailaking Task (TMT). This claim comes from the two reportedly

ignificant increases in performance for TMT B (p = 0.020) andMT difference (p = 0.004). The second of these, it should beoted, is a difference score between TMT A and TMT B ands such the analysis has violated the assumption of indepen-ence of data. Leaving that problem aside, it can be observedhat TMT A shows an impairment and so even if TMT B showedo improvement between conditions then the fact that TMT dif-erence is a difference measure means that any improvementill be partially reflecting the impairment in the TMT A task

ather than an improvement in anything. This leaves the onlyvidence for an improvement being the TMT B task. Applyinghe Bonferroni correction we would have to raise our alpha levelrom 0.05 to 0.36 in order to report this finding as statistically

ignificant.

The authors are aware of Bonferroni corrections and makeeference to it in their results. It is argued, however, that such aorrection was not appropriate given the results of a factor anal-

ologi

yaatttnarts

r

aductms

Reference

M.B. Lewis / Neuropsych

sis. A factor analysis, however, is another exploratory methodnd without further details it is impossible to see why such annalysis would negate the requirement for a Bonferroni correc-ion. The results of such a factor analysis could have been usedo further explore the data based on only two hypotheses ratherhan 18. This would have involved investigation of how the cog-itive factors (working memory and information processing) areffected by the mobile phone. For such an analysis the Bonfer-oni correction that have been a more modest division by 2 rather

han by 18. It is not possible to calculate what the results fromuch a re-analysis would be.

There is no questioning the quality and importance of theesearch reported by Keetley et al. As a matter of record and

K

a 45 (2007) 1580–1581 1581

starting point for attempted replications it is good that theirata were published but replications are required. I do find itnfortunate, however, that important and potentially emotivelaims (e.g., that mobile phones can have ‘a positive effect onasks requiring higher level cortical functioning’ p. 1847) are

ade based on levels of statistical significance far above thecientifically accepted level.

eetley, V., Wood, A. W., Spong, J., & Stough, C. (2006). Neuropsychologicalsequelae of digital mobile phone exposure in humans. Neuropsychologia,44, 1843–1848.

Recommended