kurtosis and skewness < 1. The distribution of the totals of citations is not
normally distributed, but right skewed with kurtosis = 14.7 and skewness =
3.6. The data show that there are a small number of informants with very
large numbers of citations and many with less than 2000 citations, so the
distribution more or less follows a power law.
For the two indicators, rank orders of the highest h-index and most fre-
quently cited applied linguists could be established. Table 9.4 presents the
rank orders of the 25 applied linguists with the highest h-index and the rank
order of the 25 most frequently cited applied linguists. Data for all the
applied linguists participating in this study are presented in Appendix 3.
In these tables books and articles are put together, but as we will see later
on, books attract more citations, so ideally the data should be corrected for
this. The two tables show that the two indicators used to assess impact are
both relevant. Not surprisingly, the correlations between the h-indices and
total number of citations are high (Spearman’srs=0.795;p< 0.001). Over-
all, high h-indices go with large numbers of citations, because that is what
the h-index is based on. But there are clear differences. Some researchers
have a high citation score and relatively small numbers of citations, for
example Kathleen Bardovi-Harlig and Nick Ellis. The reverse also occurs, for
example Dick Schmidt and Nina Spada. The correlation between h-index
and the most frequently cited article is 0.69, so the added value of the
second and third publication is rather small.
9.5 The impact of publications over time
When we publish an article, we hope that other researchers will pick it up
quickly and cite it. Unfortunately, that does not seem to be the way things
go. As examples I have analyzed two articles that have been cited more than
a thousand times: Norris and Ortega’s 2000 article and Firth and Wagner’s
1997 article. Figures 9.2 and 9.3 present the citations over time of both articles.
A few interesting observations can be made from thesefigures. One is that
these publications were already cited before they were published. For
Table 9.3Descriptives for h-value and total number of citations
h-value Total citationsN9898
Mean 30.08 2472
Median 29.5 1430
Mode 15 (multiple, smallest
value)
125 (multiple, smallest
value)
Standard deviation 13.38 3539
Range 66 21271
Minimum 7 125
Maximum 73 21396
The citation game 111