Assistant Data

Location:

Chapel Hill, NC

Posted:

January 03, 2013

Contact this candidate

Resume:

Library & Information Science Research ** (****) *73 281

Contents lists available at ScienceDirect

Library & Information Science Research

MPACT and citation impact: Two sides of the same scholarly coin?

Cassidy R. Sugimoto a,, Terrell G. Russell a, Lokman I. Meho b, Gary Marchionini a

School of Information and Library Science, UNC-Chapel Hill, Chapel Hill, NC 27599, USA

School of Library and Information Science, Indiana University, Bloomington, IN 47405, USA

article info abstract

This article provides the rst comparison of citation counts and mentoring impact (MPACT) indicators

Article history:

Available online 11 October 2008 indicators that serve to quantify the process of doctoral mentoring. Using a dataset of 120 library and

information science (LIS) faculty members in North America, this article examines the correlation between

MPACT indicators and citation counts. Results suggest that MPACT indicators measure something distinct

from citation counts. The article discusses these distinctions, with emphasis on differences between faculty

ranks. It considers possible explanations for weak correlations between citations and mentoring at the full

professor rank as well as implications for faculty activity analysis and broader institutional evolution.

1. Introduction counts. Comparing MPACT indicators with other measures of

productivity would allow researchers to analyze whether MPACT

Faculty members at doctoral degree-granting institutions are indicators measure something similar or distinct from existing

familiar with the three aspects on which their productivity is typically metrics. This analysis would provide valuable information for the

measured research, teaching, and service and the various metrics academic community if MPACT indicators prove to be distinct

by which their level of productivity is assessed e.g., publication measures of productivity, the academic community will have a

record, citation count and ranking, number and size of grants received, means to compare a faculty member's mentoring activity with their

number of patents issued, awards, student course evaluations, and peers', both locally and across the greater academic landscape.

level of participation and leadership in campus and professional Additionally, being able to quantify the mentoring aspect of a faculty

organizations. While each of these metrics indicates some faculty member's scholarly activities ef ciently may lead to integration

impact within their domain (discipline, eld, research area), they fail mentoring into the current academic reward system. Mentoring

to assess the mentoring impact of a faculty member adequately. could be used simultaneously with other productivity metrics.

Mentoring, speci cally in the form of advising and serving on doctoral Currently, faculty scholarly productivity assessments focus pri-

committees, is one of the most signi cant activities undertaken by marily on publication counts and venues, peer evaluation, and

faculty at doctoral granting institutions and one that requires citation-based indicators. However, these indicators of scholarship

substantial amounts of time and energy (Marchionini, Solomon, do not provide a comprehensive view of all faculty members' scholarly

Davis, & Russell, 2006, p. 482). However, this activity has been largely activities. Activities such as mentoring, which span the boundary of

overlooked in the academic community. In order to quantify this vital the academic triad integrating the ideas of teaching, research, and

aspect of scholarship, a set of mentoring impact (MPACT) indicators service should be part of a faculty member's scholarly productivity

have been proposed (Marchionini et al., 2006). These indicators assessment. If shown to provide distinct indicators of productivity,

calculate raw and weighted counts for advising and serving as a MPACT indicators may be able to re ect the consummate and

committee member on Ph.D. dissertations, resulting in values that integrated activities of the scholar accurately.

serve as proxies for the many ways in which a faculty member

in uences the eld through mentoring. 2. Literature review

1.1. Problem statement Mentoring is a multi-faceted and highly personalized form of

teaching (for de nitions, see Eby, Rhodes, & Allen, 2007). The research

The MPACT indicators have not yet been compared with other literature identi es three main mentoring situations: workplace

metrics of faculty productivity, such as publication and citation mentoring (e.g., Zachary, 2005), youth and specialized K-12 mentoring

programs (e.g., Dennis, 1993), and faculty-student mentoring (e.g.,

Daloz, 1986). This paper focuses on faculty-student mentoring. Most of

Corresponding author.

the existing research focuses on undergraduate mentoring; there is

E-mail addresses: abp4tn@r.postjobfree.com (C.R. Sugimoto), abp4tn@r.postjobfree.com

little research on advising and mentoring graduate students. In one

(T.G. Russell), abp4tn@r.postjobfree.com (L.I. Meho), abp4tn@r.postjobfree.com (G. Marchionini).

doi:10.1016/j.lisr.2008.04.005

C.R. Sugimoto et al. / Library & Information Science Research 30 (2008) 273 281

274

exception, Austin (2002) interviewed 79 doctoral students in different (e.g. citation counts, grant money, patents, publications, etc.). For

disciplines over a four-year period. He discussed the ad-hoc quality of example, the new metric begs the question, do the MPACT indicators

faculty mentoring as an important challenge for graduate education. identify something that is not already captured by an existing metric

for faculty productivity? Speci cally, how do the MPACT indicators

There are several possible reasons that graduate mentoring is not

more considered in the literature or in practice. One reason might be correlate with other existing metrics, such as citation counts, which

that just as university faculty are not expected to take instructional are increasingly important indicators of faculty research productivity?

training before assuming teaching responsibilities, they are not In an attempt to address this question, this study gathered and

expected to have any instruction in advising and mentoring. Assessing compared MPACT indicators and citation counts for a sample of 120

teaching and mentoring adequacy is a thorny issue mainly handled by LIS faculty members.

student evaluation surveys and peer observations.

This study presents doctoral-level mentoring as the rari ed epitome 3. Methodology

of university mentoring and dissertations as the best quanti able proxy

for that mentoring process. Focusing on doctoral mentoring is limiting In order to compare faculty citation counts and MPACT indicators

for the eld of LIS, a strati ed-random sample of faculty members

because it only applies to faculty at doctoral degree-granting institu-

were selected (a ten percent sample of the eld, sampled proportio-

tions. However, it is practical due to the publicly accessible record and

evidence of published work that completes the dissertation committee nately by rank) along with the entire faculty from three LIS schools.

process. The MPACT factors derived from doctoral advising and Researchers compiled their citation counts and a list of all the

committee membership are thus one practical, but limited, indicator dissertations on which they have served (see Fig. 1 for a diagram of the

of mentoring activity. Clearly, MPACT is neither necessary nor suf cient data collection and analysis process). The nal dataset included 120

as a measure of mentoring. However, it does offer one tangible kind of faculty members currently serving at 29 doctoral degree-granting

evidence for professorial mentoring in doctoral degree programs. ALA-accredited schools.

Library and information science (LIS) faculty members were the

rst group to be examined utilizing this new metric. MPACT indicators 3.1. Selection and description of faculty and schools analyzed in the study

were applied individually and in aggregate for faculty members at six

LIS schools in the U.S. (Marchionini et al., 2006). The ndings from the In order to compare individual faculty members and schools, this

study used two methods of sample selection (Phase A). The rst was to

study on the aggregate level found that for all six schools, a small

choose three schools which would be studied in their entirety that

portion of faculty members accounted for more than half of all the

doctoral dissertation mentoring that occurred, with some schools is, all faculty members from those schools were included in the

appearing slightly more distributed than others (Marchionini et al., sample. Researchers chose the University of North Carolina at Chapel

2006). In an evaluation of the indicators, the study found that the Hill (UNC), the University of Illinois at Urbana-Champaign (UIUC), and

fractional MPACT value similar to Price & Beaver's (1966) idea of Indiana University Bloomington (IUB) because complete dissertation

fractional productivity may be the most accurate method for data for these schools existed in the MPACT database. These schools

estimating mentoring productivity. It takes into account the different were also chosen because of they are ranked highly among LIS schools

roles of advising and serving as a committee member in a logically (America's best colleges, 2007), their faculty is diverse (ALISE, 2004),

weighted fashion. However, the study also found the more easily and their faculty is among the most published and cited in LIS (Adkins

calculated summation score (equally weighting advisorships and & Budd, 2006; Budd, 2000; Persson & str m, 2005). Thus, they are

committeeships) to be highly correlated with the fractional score and comparable on a number of productivity factors. In May 2007, these

thereby nearly equally useful in evaluation (Marchionini et al., 2006). schools employed a total of 61 full-time faculty members.

In order to generalize the ndings, researchers used a strati ed-

Marchionini et al. (2006) presented and evaluated the new MPACT

metric, but they did not examine the relationship of the MPACT metric random sampling method to select a group of 80 LIS faculty from ALA-

and other metrics traditionally used for assessing faculty productivity accredited programs. The sample included a proportionate number of

Fig. 1. The data collection and analysis process.

C.R. Sugimoto et al. / Library & Information Science Research 30-200*-***-***-***

the bibliographic record of any of these missed citations was found in

Table 1

Original sample before duplicate faculty and non-doctoral degree-granting schools Scopus or Web of Science, an examination was conducted to

were removed

determine why it was not retrieved and whether it should be counted

as a citation. Items that were overlooked due to searching errors were

Assistant prof. Associate prof. Full prof. Total

counted as found by their respective databases. Citations that were

Illinois faculty 3 11 6 20

Indiana faculty 5 8 5 18 missed due to errors beyond the control of the investigators were

UNC faculty 6 7 10 23

tallied but were not counted as found. Examples of these system

Random faculty 30 26 24 80

errors include records with an incomplete list of cited references, a

lack of cited references information, and errors in cited reference

information. Most of the system errors, which constituted less than 2%

of the sample's 20,486 citations, were found in Scopus. It should be

faculty from each academic rank (assistant, associate, or full professor)

emphasized that errors in cited reference information may be caused

(Table 1). This sample represents approximately 10% of the entire

by the citing authors rather than errors caused by the databases

population of full-time LIS faculty in North America (Meho, 2007).

themselves (e.g., if an author cited a speci c item and had a typo in the

From this initial list, duplicates were removed, curriculum vitae

title or author elds, this error may not be detected or corrected

(CVs) were gathered, and faculty members without direct opportu-

nities to mentor were identi ed (Phase B). Because the entire set of during data entry into the citation database).

All citations were entered into a spreadsheet and database and

full-time, tenure-track faculty of North American programs listed in

were coded by rst author, source (e.g., journal and conference name),

ALISE were sampled, there were 11 duplicate faculty members

between the three schools and the random sample. After removing document type (e.g., journal article, review article, conference paper),

publication year, language, institutional af liation of the correspon-

those 11, the selection included 130 individual faculty members.

The CV for each of these 130 individuals was gathered by means of dence author, and country of the correspondence author, as well as the

locating a current Web page or by individual solicitation via email. These source used to identify the citation. Virtually all citations were from

CVs were then mined for a list of all the schools at which the faculty refereed sources. Approximately 10% of the citations did not have

country and institutional af liation information. Researchers used the

member had taught. Since the purpose of this study was to evaluate the

relationship between citation counts and MPACT indicators, the 10 Web to identify missing information.

faculty members who had never taught at an LIS doctoral degree- Because some journal and conference names are not entered

granting institution were removed from the study. They did not have a consistently in Scopus and Web of Science (e.g., Information Research

direct opportunity to generate an MPACT score greater than zero. is indexed as Information Research in Scopus and Information Re-

search An International Electronic Journal in Web of Science), all such

The resulting list of 120 faculty members (Phase C) included 35

assistant professors, 47 associate professors, and 38 full professors instances were manually standardized. In cases where a citing source

currently serving at 29 different ALA-accredited schools in North had changed its name in the citation, the citations were merged under

America (Table 2). Appendix A lists the schools at which these 120 their most recent respective names (e.g., citations found in the Journal

faculty members currently serve. of the American Society for Information Science were listed under its

Using the CVs, researchers compiled a list of all the LIS doctoral more recent name, the Journal of the American Society for Information

degree-granting institutions at which the faculty members had Science and Technology).

taught, along with the corresponding years that they served at those

institutions. This list of 30 schools (Appendix B) represented all the 3.3. Compilation of dissertation data

schools at which the faculty member could have served on a

The compilation of dissertation data was dictated by the list of 30

dissertation (excluding service as an external member at a different

school). This list provided the focus of the study it dictated for which doctoral-degree granting schools at which any of the 120 faculty

schools researchers needed to gather dissertation data. members had served (Appendix B). This list was augmented by

identifying all years at which any of the 120 faculty members served at

3.2. Compilation of citation data the given schools. After identifying the data, researchers generated a

list of all the dissertation authors for the schools during those given

As recommended by Meho and Yang (2007) for LIS, both Scopus years (Phase D). Authors already in the existing MPACT database were

veri ed, and new authors were added using ProQuest's Dissertations

and Web of Science were used to compile citation data for the 120

sample members. The exact match search approach in Scopus was and Theses database. This was done by searching each school

used to identify citations to all items published or produced by the 120 individually with the keywords information systems or library science.

faculty members constituting the study sample. This exact search The results were then manually examined to determine whether the

method used the title of an item as a search statement (e.g., Infor- dissertation was conferred by the appropriate school/department (for

mation Seeking in Electronic Environments) to locate an exact match in example, the above search returned dissertations from education,

the cited References eld of the indexed records. In cases where the computer science, business, etc.). In some cases, this meant examining

title was too short or ambiguous to refer to the item in question, more than 500 results to add fewer than a dozen dissertations to the

additional information such as the rst author's last name was used to dataset. Also, the information regarding the school/department

ensure that researchers retrieved only relevant citations. In cases appears on the dissertation itself and not the bibliographic surrogate,

where the title was too long, the rst few words of the title were used so the exact information given varied by dissertation and school. In

because utilizing all the words in a long title may increase the some cases, the conferring department/school had to be uncovered by

possibility of missing some relevant citations due to typing or examining information on the document, such as the advisor names

indexing errors. In Web of Science, the Cited Author search option and acknowledgements. Unfortunately, ProQuest's Dissertation and

was used to identify citations to all items published or produced by the

sample.

To ensure that citations were not overlooked because of searching Table 2

or indexing errors, all of the citations missed by each database were Final faculty sample by rank

cross-examined. For example, if a citation was found in Scopus but not Assistant prof. Associate prof. Full prof. Total

in Web of Science, bibliographic searches were conducted in Web of

29 schools 35 47 38 120

Science to see if the item was in fact indexed in the database. When

C.R. Sugimoto et al. / Library & Information Science Research 30 (2008) 273 281

276

Fig. 2. Citation counts across all 120 faculty members.

Theses database does not allow searching by exact departments/ that were not available online, at UNC-Chapel Hill, or via interlibrary

disciplines, and the department names vary across the LIS discipline,. loan. In these cases, as in those in which the dissertations were unable

The researchers' method provided the most thorough way of gleaning to provide information, the schools and/or the authors were contacted

names of all authors for the given time periods. In cases where there directly. There were a few cases in which no resolution was obtained.

were signi cant problems nding complete data, researchers aug- At the time of the Marchionini et al. study (2006), the MPACT

mented the list of authors by contacting the individual schools and database contained 2400 LIS dissertations from 32 North American

requesting lists of recent doctoral graduates. schools. At the time of this writing, the MPACT database (found at

Once the list of dissertation authors had been compiled (along with http://www.ils.unc.edu/mpact/) provided dissertation information for

year conferred and school), advisor and committee member names more than 2700 LIS dissertations across 37 North American schools for

the years 1964 2007.

were investigated. The primary source for this step was also

ProQuest's Dissertation and Theses database, which allows one to

view the rst 24 pages for most dissertations conferred in the past 4. Results

13 years and those included in previous ProQuest digitization projects.

The nal dataset consisted of 120 faculty members, their rank,

Advisor and committee member names could be found in a variety of

places on the dissertation; the most common places included the schools, MPACT indicators, and citation counts (with and without self-

signature pages (some of which print the names beneath the citations), current through May 2007.

signatures and others which do not) and the acknowledgments. For

most dissertations, researchers had to cross-reference multiple places 4.1. Description

within the dissertation manually in order to disambiguate names

(some authors merely referred to the professor by last name only), Findings show that the citation counts (Fig. 2) displayed a classic

deal with lacking signature pages or indecipherable signatures, and power law distribution with the median citation count at 68.5. The

disambiguate unclear acknowledgements. In some cases, the dis- counts ranged from zero to 2287 with an average of 166.51 (Table 3).

sertation itself could not provide the information needed so the Very few faculty members (9%) had citation counts above 400, and

researchers contacted the schools and/or the authors to identify approximately 41% had fewer than 50.

further information. In a few cases, these ambiguities and/or missing MPACT (A + C) scores (Fig. 3) displayed similar characteristics to the

pieces of information were not resolved at the time of this writing. citation data, ranging from zero (0) to 70, with 70 being an extreme

For those dissertations where an electronic preview was not outlier. The average score across all 120 faculty members was 4.18,

available, the holdings available at the University of North Carolina at with 50% of faculty having a score of zero (0) or one (1) (Table 4).

Chapel Hill were searched. The print and micro che collections were In addition, researchers found that the data displayed character-

examined (more than 50% were in the micro che collection). istics of the Pareto distribution. In evaluating the A score (number of

Researchers used the same process with the same limitations as the times a faculty member served as an advisor) for each of the 120

electronic copies, trying to identify advisor and committee member faculty members, researchers found that 80% of the advisorships were

names. When the dissertation was neither online nor in the UNC- served by 18% of the faculty members.

Chapel Hill collection, the items were searched using WorldCat and

available dissertations were manually requested using interlibrary 4.1.1. Rank

loan. In total, nearly 250 interlibrary loan requests were made for this The relationship between citation counts and MPACT (A + C) by

project. However, some dissertations remained unavailable those rank is shown in Fig. 4. Of the 120 faculty member sample, 42 (35%)

Table 3

Descriptive statistics for citation counts

Min Max Std Dev Variance Mean Median Mode Quartiles

25 50 75

TOTAL (n = 120) 0 2287 305.2 93122.7 166.5 68.5 1, 16, 59 22 68.5 188.8

Assistant (n = 35) 0 336 73.8 5446.4 52.2 20 0, 1, 16 3 20 69

Associate (n = 47) 1 663 140.4 19715.6 110.7 68 14, 22, 35, 122-**-**-***

Full (n = 38) 5 2287 472.0 222769.6 340.3 174 56.8 174 338.3

UNC (n = 23) 1 1298 264.8 70122.5 178.2 92 38 92 250

IUB (n = 18) 1 1100 332.2 110337.5 248.1 148.5 53.5 148.5 242.8

UIUC (n = 20) 16 663 223.8 50072.1 219.3 107.5 42.5 107.5 413.3

Random (n = 69) 0 2287 342.4 117207.6 139.4 48 0, 1, 16, 59 16 48 108

All numbers are unique.

C.R. Sugimoto et al. / Library & Information Science Research 30-200*-***-***-***

Fig. 3. MPACT (A + C) scores across all 120 faculty members.

had an MPACT score of zero (0). This means they have neither advised MPACT (A + C) score of zero (0), whereas only 25% of faculty members

a dissertation nor sat on a dissertation committee. Of these 42 faculty at the three schools had an MPACT (A + C) score of zero (0).

members, 23 are assistant professors, 12 are associate professors, and Additionally, 15% of the faculty members at the three schools had

7 are full professors. Some of these professors are currently at schools MPACT (A + C) scores above 12, whereas only 15% of the faculty

without doctoral programs or with programs that have been initiated members in the random sample had MPACT (A + C) scores above 3.

recently. There are 47 faculty members (39%) with an MPACT between The citation counts displayed similar characteristics: 51% of the

1 and 4, inclusive: 12 assistant professors, 22 associate professors, and faculty members in the random sample had citation counts between

13 full professors. There are 31 faculty members (26%) with MPACT zero and 49, inclusive, while only 26% of the faculty members at the

scores of 5 or above. All of these are tenured faculty, and 18 are full three schools fell in that category. At the other end of the scale, only 6%

professors. No assistant professors have advised a doctoral student, of the faculty members in the random sample had citation counts of

and there are no assistant professors in the sample with 5 or more 400 or greater, while 15% of the faculty members at the three schools

committeeships. had counts in this range.

In terms of citation counts, 49 faculty members within our sample

(41%) had citation counts ranging between zero and 49. Of these, 35 4.2. Correlation

were assistant professors (all of the assistant professors within our

study), 12 were associate professors, and 2 were full professors. There The data were imported into SPSS to perform a correlational

were 27 faculty members (22.5%) with citation counts from 50 to 99: analysis between the MPACT indicators and full citations counts. The

22 associate professors and 5 full professors. Thirty-three faculty analysis had two separate goals: 1) to analyze the impact of faculty

members (28%) had citation counts from 100 to 399: 12 associate rank on the entire data set and 2) to analyze the patterns across the

professors and 21 full professors. Lastly, 11 faculty members (9%) had three individual schools.

citation counts of 400 or more. There was one associate professor in

this group, and the remaining 10 were full professors. 4.2.1. Rank

The rst correlational analysis involved faculty rank and all MPACT

4.1.2. Analysis by school indicators from the MPACT database. The MPACT indicators consist of

two raw components and ve derived values. The two raw components

Analyzing the dataset by school indicated that faculty members in

the three individual schools (UNC, UIUC, and IUB) had higher MPACT are A (the number of times a particular faculty member served as a

(A + C) scores (Fig. 5) than those in the random sample. This is mostly dissertation advisor) and C (the number of times a particular faculty

due to a higher proportion of tenured faculty within the three-school member served on a dissertation committee, not as an advisor). A + C is

sample the three individual schools did not have a proportionate simply the sum of these two components, weighting each component

number of assistant professors to the entire population. Because equally. The FM indicators represent fractional mentorship calculated

strati ed sampling was used to select the random sample, the as A + (1 / number of committee members). The number of committee

proportions of faculty at each rank are representative of the entire members can be inclusive (FMI) or exclusive (FME) of the number of

population (805 faculty members in all of LIS). However, the rank advisors on that dissertation committee because some dissertations

distribution in the three-school sample is skewed toward the associate include co-advisors. FMI / (A + C) and FME / (A + C) are normalized

and full professor ranks. versions of the FM indicators.

The distributions of MPACT scores by school re ected this The correlational analysis revealed primarily insigni cant positive

disparity: 42% of faculty members in the random sample had an correlations between MPACT indicators and full citation counts

Table 4

Descriptive statistics for MPACT (A + C)

Min Max Std Dev Variance Mean Median Mode Quartiles

25 50 75

TOTAL (n = 120) 0 70 8.42 70.97 4.18 1.5 0 0 1.5 5

Assistant (n = 35) 0 3 0.95 0.89 0.60 0 0 0 0 1

Associate (n = 47) 0 30 4.89 23.82 3.30 2 0 0 2 5

Full (n = 38) 0 70 12.85 165.07 8.55 3.5 0 1 3.5 10.3

UNC (n = 23) 0 30 9.75 94.98 9.39 6 0 1 6 15

IUB (n = 18) 0 18 5.38 28.89 3.78 2 2 1 2 4

UIUC (n = 20) 0 70 15.26 232.87 6.35 3.5 0 0 3.5 5.8

Random (n = 69) 0 28 4.01 16.04 2.07 1 0 0 1 2

C.R. Sugimoto et al. / Library & Information Science Research 30 (2008) 273 281

278

Fig. 4. Relationship between citation counts and MPACT (A + C) by rank.

(Table 5). Signi cant correlations were found at the associate level for were weak (.268). The most surprising result was the almost random

the A indicator (.526) and also for those non-normalized FM indicators relationship between citation count and MPACT at the full professor

that give the A indicator greater weight. The negative correlation rank; these results are discussed in Section 4.

between the A indicator and assistant professor citation counts Correlational analysis for citation counts without self-citations

( .022) re ects the fact that no assistants in the sample had served as (Table 6) re ected the same patterns as the full citation count analysis

advisor on a dissertation. Thus, the correlation is necessarily less than (Table 5). The difference between these correlations is minimal; it

or equal to zero. The relatively small negative correlation suggests that reinforces the interchangeable use of citation counts with or without

citation counts for assistant professors vary considerably. If most self-citations (Cronin & Meho, 2006).

assistant professors had very small citation counts, there would be a Recognizing the individual with an MPACT score of 70 and the

stronger negative linear relationship. Signi cant correlations were individual with a citation count of 2287 as outliers (both full

also identi ed for the whole population. This was largely due to the professors), the correlational analysis was rerun without outliers to

strength of the correlation of the associate A indictor, as well as the identify if there were any considerable changes. Table 7 shows that the

lack of assistant professors serving as advisors. However, even these correlations became more positive with the removal of the two

Fig. 5. Relationship between citation counts and MPACT (A + C) by school and random sample.

C.R. Sugimoto et al. / Library & Information Science Research 30-200*-***-***-***

Table 5 Table 7

Correlations between MPACT indicators and full citation count by rank level Correlations between MPACT indicators and citations removing the two outliers

A C A+C FMI FME FMI/(A + C) FME/(A + C) A C A+C FMI FME FMI/(A + C) FME/(A + C)

.022

Assistant .088 .081 .039 .044 .079 .083 Full .206 .045 .106 .172 .164 .007 .029

.526 .445 .431 .327 .299 .292

Associate .207 .298 .331 .324 Tenured .159 .226 .124 .136

.377 .290 .357 .352

Full .101 .012 .049 .074 .069 .080 .090 ALL .225 .216 .227

Tenured .219 .120 .154 .192 .186 .146 .155

Correlation is signi cant at the .01 level (2-tailed).

.268 .244 .239

ALL .175 .209 .227 .234

Correlation is signi cant at the .01 level (2-tailed).

practices that preclude assistant professors from serving as advisors or

serving without a tenured co-advisor. Second, assistant professors are

outliers, but only the category of tenured professors showed a change

often encouraged to focus on establishing their research agendas and

in levels of signi cance. This may be due to the higher proportion of

teaching repertoire in their rst years rather than serving on any

associate professors in that category with the removal of two full

committees. Third, assistant professors typically have not established

professors.

research agendas, funding support, or reputations that tend to attract

doctoral students. Fourth, the doctoral process takes several years,

4.2.2. School

sometimes longer than the term of an assistant professorship.

Correlational analysis was then performed on each school, to

Assistant professors may begin working with a doctoral student and

analyze patterns across sets of faculty members. Table 8 displays the

be promoted to associate professor by the time the dissertation is

correlations between MPACT indicators and full citations counts for

completed. Finally, mentoring is generally not explicitly de ned or

each school and for the entire data set.

rewarded. These conditions are somewhat mitigated by eagerness to

Correlational analysis at the school level displays high levels of

work with young scholars who might collaborate on research. Thus,

variance due to the small sample size and outliers in the data set. For

the results of this study re ect current academic practice and culture.

example, the UNC data seems to re ect the presence of an individual

It will be interesting to see whether doctoral committee service is

who is both extremely well-cited and has an unusually high MPACT

more prevalent for faculty who work in rapidly changing elds like

value. IUB also contains an outlier an individual who is extremely

popular culture or technology. In those elds, there are strong

well-cited, but has never served as an advisor on a doctoral

incentives for faculty to work with knowledgeable young people or

committee. The outlier for UIUC is also well-cited and has an

to work on highly interdisciplinary problems that require large teams.

extremely high MPACT value (A + C of 70). With the removal of these

The intuition is that dissertation committee patterns may also be used

outliers, the data changes substantially (Table 9).

as evidence for tracking research trends in schools or elds.

The results are insigni cant with the removal of the outlier from

Associate professors displayed the highest correlation between

each school, which may be partly due to the high concentration of

MPACT values and citation counts. These positive correlations show

tenured faculty members at these schools (see Table 1 and Section

that these professors are simultaneously mentoring and generating

3.1). Furthermore, data from Tables 8 and 9 may indicate that the

citations (many of which are citations of work published in their pre-

schools' samples are too small and volatile from which to draw

tenure assistant professor phase). This characterizes the surge of the

generalizable conclusions and that the combined data set provides

academic life cycle: the products created during the pre-tenure

more stable generalizations.

promotion now reward the individual with academic dividends in the

form of citations, awards, grants, students, and overall visibility within

5. Discussion

their discipline. The associate professor rank is pivotal in the academic

career. The reward of promotion (and typically tenure) provides these

The greatest good you can do for another is not just to share your dividends and also brings new demands and opportunities. The

riches but to reveal to him his own. demands are both internal and external. Psychologically, success and

Benjamin Disraeli well-honed work habits may motivate some people to accelerate their

self-expectations, decide to maintain extant inertia, or in some cases,

The data from this study shows that for the most part, there is decelerate self-expectations. Likewise, progress on complex research

little to no correlation between MPACT values and citation counts problems may stimulate opportunity overload for research direc-

implying that MPACT is measuring something distinct from citation tions, which could stimulate or confuse continued progress. In some

counts. The strongest positive correlations were at the associate level, cases, if narrow research agendas were adopted, the research veins

suggesting that there might be some relationship between the may be mined out and new directions sought. Other internal factors

academic life cycle and MPACT values. The academic ranking system include the aging process and social obligations that come with age.

(assistant, associate, full professors) seems to be related to mentoring External demands that face associate professors come from institu-

activity. tional as well as professional practice sources. Committee service

As the data show, very few assistant professors are engaged in expectations at campus and professional society levels become

dissertation mentoring. There are several possible explanations. First, greater, as do expectations about participation in critical social

to protect student advising continuity, some schools have policies or practices such as peer review. In general, the reputation associated

Table 6 Table 8

Correlations between MPACT indicators and citation counts without self-citations Correlations between MPACT indicators and full citation count at the school level

A C A+C FMI FME FMI/(A + C) FME/(A + C) A C A+C FMI FME FMI/(A + C) FME/(A + C)

.024 .759 .536 .703 .692

Assistant .082 .076 .034 .038 .070 .074 UNC .376 .342 .335

.525 .444 .430 .037

Associate .206 .297 .328 .321 IUB .066 .038 .003 .013 .065 .152

Full .103 .012 .041 .076 .070 .080 .085 UIUC .322 .238 .264 .292 .286 .375 .343

.403 .337 .385 .403 .402

Tenured .220 .118 .154 .192 .186 .145 .150 Random .219 .213

.267 .243 .237 .268 .244 .239

ALL .172 .207 .224 .228 ALL .175 .209 .227 .234

Correlation is signi cant at the .01 level (2-tailed). Correlation is signi cant at the .01 level (2-tailed).

C.R. Sugimoto et al. / Library & Information Science Research 30 (2008) 273 281

280

collaboration between people who aim to grow and learn. Classically,

Table 9

Correlations between MPACT indicators and full citation count at the school level (with there is more initiation and control by the mentor rather than the

outliers removed)

mentee. However, as a mentoring relationship matures, the contribu-

tions become more balanced and mentors learn and evolve too. As

A C A+C FMI FME FMI/(A + C) FME/(A + C)

Kram's (1985) model of mentoring demonstrates, mentoring activities

UNC .286 .439 .432 .388 .401 .219 .241

IUB .067 .077 .079 .083 .089 .189 .285 are as much about psychosocial exchange as content and career advice.

UIUC .336 .042 .183 .278 .257 .353 .318

Mentoring must be done in a context: topics of discussion range

beyond the topic of interest to ideas, feelings, and events central to the

respective people. Mentoring takes time time to develop trust and

establish common communication patterns. Admittedly, this is a

highly idealized notion of mentoring, but most successful adults can

with the research, teaching, and service done to achieve promotion

point to one or more mentors that in uenced their lives.

brings more research opportunities, students, and service invitations.

Although the relationship between dissertation advisor and

Associate professors make implicit and explicit decisions about what

student is not always one that involves even a few aspects of

kind of career trajectory they wish to pursue based on their

experiences during assistant professor boot camp and their inherent mentorship, it stands as the most consistent and public exemplar of

mentoring in the academy. Thus, MPACT factors are one way to

personal compositions.

measure how much mentoring a professor does over a career. It is

Once the individual has obtained the status of full professor, these

crucial to qualify MPACT analysis with the recognition that disserta-

data show that the correlation between MPACT values and citation

tion advising or committeeship are neither necessary nor suf cient to

counts is lost. This may indicate that personal decisions become more

of a factor in this post-promotion phase some choose to mentor, good mentors and only represent one kind of mentoring that faculty

some choose to do research, some choose to do both, and some do do. However, MPACT is a strong indication of mentoring and it is

neither. This is surprising and somewhat disturbing. It is essential to

Contact this candidate