[Contents Page] [Previous section] [Next section] Back to my homepage
See Tables 15 (serials) and 16 (non-serials). The sample size is small, but some patterns seem to emerge for the n.h. periodicals (Table 15):
It must be emphasized that these patterns probably arise less from the nature of the subjects or the library's holdings than from the sample size and the particular thesis topics (e.g., Ed09 & Ed10, Ps01 & Ps13, etc).
The list of "popular" n.h. non-serial items is astonishingly small, and seems to be a product of the closeness of the thesis topics, or perhaps represents resources for courses taken together by the various Education students. All 5 items are dated and not even frequently used by their citing theses. Therefore none of the 5 are recommended for acquisition. Anomalously, 2 of the 5 are "foreign" Education theses, which is out of all proportion to the use of thesis citations or interloans --- suggesting that they may be the personal property of a member of staff commonly known to the thesis writers.
The citation study cumulative distributions are given in Table 19.[133] Note that although the subject curves are somewhat cluttered (Table 11), there is a clear distinction between the H and SS cumulative curves (Table 12).[134]
If the citations were truly exponential and the printed and non-printed matter distributed in the same proportions, through time, the mean citation age would be inversely proportional to the rate constant. Therefore we expect the cumulative curves to fan out, approximately in accord with their mean citation ages.
Indeed History and English do clearly separate from the SS , and History (which has the greater mean citation age) is the more gradually accumulated. However Psychology reaches 100 % unusually early, and the large sample size would suggest that this is a significant deviation.
By a log-linear plot of the complements of two curves (see Graphs 14 and 15), we may deduce the half-life of the H and SS populations --- namely 20.9 and 9.51 years, respectively.[135] Obviously the half-life only has direct relevance to the "downward" slope of y(t) or u(t), i.e., after t*.
Assuming that we wish to retain about 98 % of the generic printed stock, as measured by demand: the default cut-off age for SS material is 39 years;[136] or 32 years for Psychology, but 50 years for Education. Given that Anthropology and Political Studies had such a small sample, their rates could be significantly distorted (up or down) by the various topics successfully presented in 1996. Since their works obsolesce faster that the general SS rate, it is wiser to use the SS rate in the absence of other evidence.
Conversely the H cumulative table does not even reach 98 % in 96 years. An approximate value of 103 years may be extrapolated;[137] However the small sample size (5 theses each in only 2 subjects) makes such an estimate dubious.
The reference study data is similar, but generally at a slower rate, confirming suspicions that each older item is used less intensively than each recent item.
There is a distinct difference in the comparative availability of resources for each subject; but it must be emphasized that the sample of theses, and hence the topics within each subject, is small. Furthermore the choice of a particular topic may seriously effect whether the library might be expected to hold relevant material.[138]
As may be seen from Table 13, the percentage of h. items, by subject, varies from 70.6 % to 86.0 %; the n.h. items are 14.0 % to 25.2 %; these figures do not complete agree because of the various unidentifiable items, namely:[139]
| English | 0.00 % |
| Political Studies | 0.00 % |
| History | 0.20 % |
| Anthropology | 1.78 % |
| Education | 3.28 % |
| Psychology | 5.22 % |
It is significant that the H have almost no unidentifiable citations (2 total).[140] This is attributed to the H preference for Harvard A, or Chicago A, style citations --- thus even if the cited work is omitted from the bibliography, enough details are given that identify the work and its formats.
Conversely, the citation practices evident in the Education theses (in particular) were markedly poorer. Some had negligible inconsistencies; but several were woefully incompetent or erroneous.
The two subjects with no unknown items had the highest percentage of h. items; but this is probably coincidence, because History (almost no unidentifiable items) had the highest percentage of n.h. items. It is unknown whether there is any connection between the high holding rates for English & Political Studies and the fact that all their citations were in English.
Similarly, the high holding rates for English and Political Studies superficially indicate that these subjects are well covered by the library; but the answer may also be that these students are more capable of adapting their topic, or how they attack it, based upon the library's actual resources. Given the consistency of their citations, it may well be that these candidates are more capable of using the library system.[141]
The consistently identifiable nature of the English, History, and Political Studies citations probably reflects a strong emphasis on good citation practice in the respective Departments, and the small number of theses (i.e., 5 from each) made it statistically unlikely that even one thesis that cited poorly would slip through. Conversely several of the 16 Psychology theses had excellent or perfect citation consistency, but this is not perceivable in the subject summary (5.22 % of all citations were unascertainable).
The conceptual formats of the citations are summarized on each subject distributions (Tables 1-6). See Table 13 for the analysis of the citations cross-referenced by known format. Again, missing percentages represent items of unknown format.
Some of the results were reasonably expected (e.g., Education was the only subject that used Eric documents). Here are some of the highlights of Tables 1-6 and 13:
To viably compare serial usage with previous published results,[149] we generally require the reference study data (1A-6A, 13A):
| Subject | Cit.Study | Ref.Study | D.& K. |
| Psychology | 66.14 % | 63.84 % | 35.0 |
| Anthropology | 35.91 % | 45.05 % | 42.9 |
| Education | 35.09 % | 36.42 % | 40.5; 41.7 |
| Political Studies | 27.36 % | 26.96 % | 31.5; 34.6 |
| History | 25.35 % | 27.73 % | 23.3 |
| English | 24.84 % | 20.55 % | 13.3; 19.9; 28.1 |
The percentages are included in Table 13. The items themselves are given in Table 14. Note that for all these items, they are only used in one thesis each; none makes the popular lists. Furthermore only one thesis in each subject is responsible for all the citations for that combination of language and subject. Two subjects, English and Political Studies, have no foreign language citations. This apparently wide-spread lack of a second-language is surprising.
All the use of primary sources in a foreign language is traceable to thesis Hs4, except for the 2 French references of Ps08.[150]
These are tabulated in Tables 17 (serials) and 18 (non-serials). Many of these titles are indeed familiar, although some of the lower ranked items are only listed because of the closeness of particular thesis topics.
For the first 63 h. serials (i.e., all items used by 3 or more theses), only 2 are in non-print form: the N.Z. Herald and the Wellington Dominion. All the h. non-serials are also in print format, except for two educational microfiche items.
It is evident, considering the number of theses and popular h. non-serials, that the average Education thesis is disproportionately likely to cite a source used by an other [Education] thesis. This is only partially attributable to a commonality of various thesis topics. Comparing the figures for Education (20 theses) and Psychology (16 theses), two other possible explanations are:
One apparent answer is that Education theses average 233 citations (while Psychology theses average only 174 citations) --- therefore, by firing more shots, the Education theses are bound to hit more targets. But that argument, actually, is false. The reference study establishes that the average Education bibliography contains 83 items that are actually used, and the average Psychology thesis has 76; the figures are too close to explain the significant number of shared Education resources.
One final point of interest. The mean age of h. non-serial items is 9.8 years (Deduced from Table 18) which under both the Education mean age of 10.91 years, or the SS mean age of 10.95 years.[151] Is recency (or at least being more recent than the average resource) a significant factor in deciding the popularity of a non-serial?
133. The corresponding reference data is in Table 19A
134. Furthermore, the SS curve is much smoother, but this is not unexpected because the SS sample is much larger (by a factor of 6.17).
135. To avoid the obvious errors arising when the citation age is small and y(t), or u(t), deviates from the exponential envelope discussed in §2.6 above, each log-linear regression line was deduced only for citation ages greater than 1.5 time the experimentally derived mean citation age for that distribution --- i.e., only t>21 (for H curve) and t>15 (for SS curve) --- when it might be assumed that each u(t) approximates a simple decaying exponential function. It may be readily shown that, if B is the slope of such a regression line (where the complement is expressed as a number between 1 and 0, and the natural logarithm is used), the half-life is given by L = - ln(2) / B
136. As per Table 19.
137. I.e., extrapolated from the 96 % value of 82 years, given a half-life of 20.9 years.
138. E.g., Ap3 clearly draws on a specialist collection to which the candidate had personal access. Most of Ap3's n.h. items would not be considered relevant to the University's allocation of Anthropology materials, so one is not surprised that so many are n.h.; yet this one thesis represents 20 % of the 1996 thesis topics.
139. Of course the figures for the reference study are even more alarming, respectively:
0.00 % --- 0.00 % --- 0.39 % --- 4.50 % --- 5.71 % --- 7.17 %
140. Or 0 % of English citations and 0.20 % of the History citations. See Tables 1 and 2.
141. Unfortunately these suppositions require verification from other sources --- e.g., an appropriate survey.
142. Approximately equal to Political Studies.
143. 2.0% h., 1.2% n.h.; which approximately equals Anthropology's total.
144. But note the comparatively high n.h. figure (1.6 %) --- cf. Anthropology and Education.
145. 2.6 h., 0.6% n.h.; which approximately equals History's total.
146. No H use electronic resources; but all SS do.
147. Education has negligible use on a monographic videotape (0.02 %).
148. An other exception which English and Political Studies alone share.
149. In this case, the figure summarized in Devin and Kellogg, "Serial/Monograph Ratio," Table 3, 50. But see also Bowman, "Format Citation Patterns," 4-5, for a generally more detailed breakdown of the formats.
150. These are numerically implicit [n-type] primary sources, but examination of the bibliography of Ps08 shows that both these items are indeed central to the topic of the thesis.
151. We must use the reference study figures from Tables 4A and 8A, because the dates in Table 18 have been given weighting by the number of theses using them --- i.e. by the number of references, rather than the number of citations. The average age of "popular" n.h. non-serials is much greater, namely 15.6 years.