how study design influences the ranking of medicinal plant...

How Study Design Influences the Ranking of Medicinal PlantImportance: A Case Study from Ghana, West Africa

KATHERINE DUDNEY*,1, SARAH WARREN2, ERIN SILLS2, AND JERRY JACKA3

1AECOM, 1333 Broadway, Suite 800, Oakland, CA 94612, USA2Department of Forestry and Environmental Resources, North Carolina State University, 2800 FaucetteDrive, Campus Box 8008, Raleigh, NC 27695, USA3Department of Anthropology, University of Colorado, Boulder, 1350 Pleasant St., Campus Box 233UCB, Boulder, CO 80309-0233, USA*Corresponding author; e-mail: [email protected]

How Study Design Influences the Ranking of Medicinal Plant Importance: A Case Study fromGhana,West Africa. Indices that rank medicinal plants by their cultural importance are valuable tools bothfor developing conservation priorities and locating effective medicinal compounds. Previous studies havecompared different indices, focusing on the formulae used to construct them. The final cultural importanceranking can also be affected by other aspects of the research methods, such as selection of informants anddesign of the data collection protocol. We assess the influence of these different methodological choices bycomparing rankings of medicinal plants developed with ethnobotanical data collected in five rural villages inGhana in 2006. We compare six indices of importance, five that are commonly used and one newlydeveloped based on responses from different sub-samples of our informants. Overall and within each sub-sample, we find little difference in plant rankings suggested by indices constructed with different formulae.Likewise, we find few differences in rankings based on different aggregations of use values. The moresignificantly influential methodological choices seem to be whether or not to prompt informants bymentioning specific illnesses, and which informants to interview. Specifically, we confirm the commonbelief that older respondents list different medicinal plant species than younger generations, and that menand women cite different medicinal species. In sum, our findings suggest that conservation priorities arelikely to be relatively robust to the particular index and use categories employed to rank plant importance,but sample selection and elicitation methods may significantly influence results.

Key Words: Traditional ethnobotanical knowledge (TEK), Ghana, medicinal plants, quantitative eth-nobotany, informant consensus, cultural importance.

Introduction

Identifying the plant species that are most impor-tant to local populations is a common objective ofethnobotanical research (Heinrich et al. 1998;

Kristensen and Lykke 2003). To provide standard-ized, replicable analyses, numerous indices havebeen developed for quantifying plant importance(Gomez-Beloz 2002; Kvist et al. 1995; Phillipsand Gentry 1993a; Reyes-García et al. 2006; Silvaet al. 2006; Trotter and Logan 1986). These indicesrequire different data collection methods (e.g., fieldidentification, free-response interviews) and baseimportance rankings on different response or spe-cies attributes (e.g., the number of times reported,number of uses, respondent consensus). Buildingon previous studies that have compared rankingresults from different indices (Reyes-García et al.2006; Tardío and Pardo-de-Santayana 2008), we

1Received 12 April 2015; accepted 16October 2015;published online 17 December 2015

Electronic supplementary material The online versionof this article (doi:10.1007/s12231-015-9322-y) containssupplementary material, which is available to authorizedusers.

Economic Botany, 69(4), 2015, pp. 306–317© 2015, by The New York Botanical Garden Press, Bronx, NY 10458-5126 U.S.A.

compare rankings of medicinal plants in Ghanabased on six different formulae for constructingindices. We then further contribute to the literatureby subdividing the data to assess the influence ofdata collection methods on the results.

Five of the indices (frequency cited, uses totaled,cultural value index, relative importance index, andinformant consensus index) compared in this studyhave been previously used to quantify plant impor-tance and indigenous knowledge. Many variationsof the informant consensus method (Trotter andLogan 1986) have been described in the literature(Amiguet et al. 2005; Begossi et al. 2002; Case et al.2005; Gomez-Beloz 2002; Hanazaki et al. 2000;Heinrich et al. 1998; Kristensen and Lykke 2003;Phillips and Gentry 1993a, b; Reyes-García et al.2006; Voeks and Leony 2004).We develop and usea logarithmic informant consensus index as oursixth index. While our comparison includes someof the most commonly used indices, it is not com-prehensive: there are both other indices, such asthose to identify conservation priority (Kala et al.2004; Martinéz et al. 2006), and other quantitativemethods for measuring importance, including use-value (Prance et al. 1987) and relative importance(Albuquerque et al. 2006).

Prior to analyzing importance by any of theseindices, a researcher must make critical decisionsregarding how the plant use data are collected. Thesample of respondents may be random, stratified, orspecifically targeted from within a culture, a village,or a subset of the population. For studiesdocumenting medicinal plant usage and properties,identified herbalists or village elders are often targetedfor interviews due to their presumed higher level ofknowledge (Johns et al. 1990). Some researchersprovide plant specimens to their respondents or taketheir respondents into the field to elicit informationabout a specific plant’s uses (Philips et al. 1994;Prance et al. 1987). Other researchers conduct in-terviews using open-ended questionnaires that allowrespondents to reply freely, naming any plant thatthey can recall during the interview (Johns et al.1990). Information collected during prompted in-terviews (where respondents are asked about specificplants or uses) may result in more consistent andspecific data, but is susceptible to researcher bias inthe selection of plants and uses; free-response ques-tionnaires may be less subject to researcher bias andallow for more diversity in the set of responses. Thisstudy considers the effects on importance rankingresults of both respondent selection (i.e., the influ-ence of respondent demographics) and survey

administration (i.e., prompting during free-listinginterviews).

In addition to interview structure and data col-lection, researchers control how data will be aggre-gated for analysis. For example, medicinal uses maybe separated into specific uses (e.g., the treatment ofboils, the treatment of measles, rashes) or lumpedinto broader categories (e.g., treatment of dermato-logical ailments). If separated into specific uses, aplant’s value might be overestimated by assigningtoo much weight to similar uses (e.g., rash andmeasles). Likewise, a plant’s value might beunderestimated if uses are lumped too broadly andspecialized uses are combined with more generaluses (Gausset 2004; Hoffman and Gallaher 2007).Data on plants can be analyzed by the part of theplant used in the treatment, the species of plant, or abroader group of plants (e.g., taxonomic family).

The selection of who to interview, how to ad-minister the questionnaire, and which index to usewill depend largely on the research questions. If theaim is to document traditional knowledge that is atrisk of being lost, a researcher may want to interviewherbalists and elders using a free-listing question-naire. If trying to identify effective medicinal com-pounds, a structured questionnaire about specificspecies analyzed using an informant consensus in-dex might be most appropriate. If trying to identifyspecies for conservation, the number of uses, fre-quency cited, or use value along with an estimationof threat or harvest technique would provide theresearcher with a metric for conservation value.These are just a few of the considerations thatresearchers consider in designing their studies.

In this paper, we use Spearman rank correlationcoefficients to compare rankings and to address thefollowing research questions:

1) Are importance ranking results correlated acrossindices?

2) Are correlation patterns across indices the sameregardless of how similar uses are grouped intouse categories?

3) Are the correlation patterns observed for theentire dataset also observed for subsets of thedata based on either respondent demographicsor interview structure (i.e., prompted vs. un-prompted interviews)?

By answering these questions, we gain insightsinto how the index formula, use categorization, anddata collection protocol influence the resulting im-portance rankings. If ranking results are highly

307DUDNEY ET AL.: IMPORTANCE RANKING COMPARISON2015]

correlated in our case study from Ghana, we couldpresume that these factors do not necessarily affectrankings. If the ranking results are not highly cor-related, then we could identify methodologicalchoices that can influence the resulting ranking.

Methods

STUDY SITE AND RESPONDENTS

The study site is located near Mount Afadjato,the tallest mountain in Ghana (Fig. 1). Due to itselevation, the site is a refuge for moist deciduousspecies normally only found at lower latitudes and isrecognized for its high levels of biodiversity (Ayensuet al. 1996).Five communities surrounding Mount Afadjato

(Fig. 1) were surveyed to assess their knowledge ofmedicinal plant species. These communities, work-ing together with assistance from the Ghana Wild-life Society, have established a 12 km2 forest reserve,have developed small-scale enterprises, and are pro-moting local ecotourism. The sample included threevillages (400–500 residents each) located at the footof the mountain range along the main road. Theseresidents typically owned land and worked in arange of professions as farmers, teachers, craftsmen,barbers, sales people, forest guards, religious offi-cials, and healers. The survey was also conductedin two settlement communities of tenant farmers(less than 100 residents each) located at the top ofthe mountain range; unlike the three villages, thetenant farmers lacked easy access to roads or elec-tricity. Village residents, due to their relativelyhigher incomes and easier access to markets, tendedto rely more heavily on Western and purchasedmedicines, while settlement residents were morelikely to collect and make their own herbal remedies(Caldwell 2007).A semistructured open-ended survey of medici-

nal plant knowledge was developed using expertreview and pretesting. During the pretesting, 20interviews were conducted in order to finalize theinterview questions. At this time, it was noted thattreatments for seven common illnesses were repeat-edly mentioned by respondents, likely due to theprevalence of these illnesses within the community.These seven illnesses, fever (typically associated withmalaria), stomach problems, blood (anemia), cold/cough, headache/body pain, rashes, and cuts, wereeventually selected to be used as interview promptswhen respondents had exhausted their free-listing

recall. Referred to herein as Bprompted^ responses,mention of these treatments served to facilitate thediscussion and remind respondents about moremedicinal species. In June and July 2006, 214interviews were conducted by the lead author withassistance from local interpreters. All respondentswere selected through convenience sampling, al-though interview times were deliberately spreadthroughout the day in order to obtain a representa-tive distribution of age groups and genders. Approx-imately 20–35% of the adult population in eachcommunity was surveyed.The survey collected information on demo-

graphics, medicinal plant use, and medicinalplant preferences. The open-ended questionsasked respondents to free-list all of the medicinalplants with which they were familiar (responsesherein referred to as Bunprompted responses^).Once a respondent’s recall was exhausted, any ofthe seven prompt illnesses that had not alreadybeen mentioned were suggested to solicit addi-tional information. For each plant named, infor-mation was collected about the preparation, ill-nesses treated, and dosage. Additional questionselicited the respondent’s views on conservation,the availability of species (now and in the past),medicinal knowledge in the community, andpathways for knowledge transmission.Respondents were approximately uniformly dis-

tributed by sex and across three age cohorts (18–40,41–60, or 61+). Education approximated a normaldistribution, with an average of middle schooleducation.Most interviews were conducted in Ewe, the local

language. Since much of the published literatureonly lists common names of the language of themajority tribe (Akan), numerous sources wereconsulted to identify scientific names of species.Several previous studies conducted in the area bythe Ghana Wildlife Society were immensely valu-able for linking local and scientific names (e.g.,Kpordugbe 2000; Owusu 2001; Swaine et al.1999; Tettey et al. 2005). Scientific names andinformation about the species were also collectedfrom labeled plants, signs, and literature at theAburi Botanical Gardens, at theDr. Noamesi Herb-al Clinic in Hohoe, the District capital, previouslypublished books (Abbiw 1990; Ayensu 1978;Hawthorne 1990; Iwu 1993), and from local ex-perts (e.g., reserve forest guards and local herbalistsas noted in acknowledgments) (Caldwell 2007).Plant specimens were collected with the help offorest guards and provided to the Ghana Wildlife

308 ECONOMIC BOTANY [VOL 69

Society. Scientific nomenclature follows the recom-mended format of The Plant List (2013).

QUANTITATIVE ANALYSES

The interview responses were used to con-struct six indices of plant importance. Whilethere are numerous indices used in ethnobot-any, these particular indices were selectedbased on the type and method of data collec-tion (i.e., specific information on medicinalplants collected by free-listing). In addition,the authors identified and selected indices thatemphasized value in distinct ways (e.g., num-ber of times cited, number of uses, consistencyin use). As such, some commonly used indices,such as use value (Rossato et al. 1999) andrelative importance (Bennett and Prance 2000),were not employed because they were either toohighly correlated with the other methods selectedor they were not appropriate given the type of data

available and research questions. The following in-dices were used to rank importance for this study:

Frequency Cited (Freq.): Frequency cited is thenumber of times a plant is cited for any medicinalpurpose (also called a use report) (Bonet andVallès 2003). This index ranks species high iffrequently reported, regardless of the number ofuses or consensus among those uses.

Uses Totaled (# Uses, Count, Researcher Tally):Uses totaled is the total number of uses reported.This index does not weight the importance ofdifferent uses; rather, it assigns the highest valueto the species with the most uses (Hoffman andGallaher 2007). Thus, a plant may be cited manytimes because of its availability or cultural impor-tance, but that number does not necessarily in-dicate whether the species contains highly effec-tive medicinal compounds (Trotter and Logan1986).

Fig. 1. Study area map. Villages and settlements surveyed in this study are circled.


Cultural Value Index (CVs): This method wasdeveloped as a function of both the number ofuses and the number of respondents reportingthe plant (Reyes-García et al. 2006). First, thenumber of uses of a species (NU) is divided bythe total number of use categories (NC). Second,the number of respondents who listed a species asuseful (R) is divided by the total number ofrespondents (N). Third, the sum of the frequen-cy cited for each use is divided by the number ofrespondents. These three parts are multiplied toprovide the cultural value index. This indexvalues species that have a high number of usesand high number of respondents, but does notvalue consensus among respondents.

CVs ¼ NUs=NC� �� Rs=N

��

X

u

FCus

N

2

664

3

775

2

664

Relative Importance Index (RI): Like culturalvalue index, this index was developed toproduce rankings influenced by both thefrequency of citation and the number ofuses (Pardo-de-Santayana 2003; Tardío andPardo-de-Santayana 2008). It averages the max-imum frequency cited for the study (RFC) andthe maximum number of uses for all speciesstudied (RNU). This index is not influenced bythe number of respondents or consensus amonguses. It may bias rankings towards species that arefrequently used, but not necessarily economicallyvaluable or medicinally effective.

RI s ¼ RFCs þ RNUs

2

Informant Consensus (IAS): The informant con-sensus method was developed to identify poten-tially effective medicinal plants as indicated by ahigh level of consensus among respondents(Trotter and Logan 1986). It has been influentialin quantitative ethnobotany studies (Hoffmanand Gallaher 2007) and has been adapted bymany researchers (Gomez-Beloz 2002; Heinrichet al. 1998; Kremen et al. 1998; Kvist et al. 1995;La Torre-Cuadros and Islebe 2003; Lykke et al.

2004; Reyes-García et al. 2005). Informant con-sensus in its simplest form subtracts the numberof uses for a species from the number of reports ofthat species and divides the result by the numberof reports for the species minus one. Informantconsensus has been cr i t ic ized for notdistinguishing degrees of importance and forranking rarely used species with high consensusover species that are popularly used for manydifferent purposes (Hoffman and Gallaher2007; Kvist et al. 1995).

ICs ¼ FCs−NUs

FCs−1Logarithmic Informant Consensus (LIC): To ad-dress criticisms of the informant consensus meth-od, we developed a modified method that takesinto account both frequency of use andinformant consensus. This method provides aconsensus value for each use, similar to whatTrotter and Logan (1986) developed for eachspecies. This value is then multiplied by thenatural log of the number of reports of a speciesuse for a species. The result is summed across alluses to obtain the logarithmic informant consen-sus score for each species.

LICs ¼X

u

ICu* ln FCusð Þð Þ

Where:

ICu ¼ FCu−NSu

FCu−1

The objective of this method is to obtain a con-sensus index that is also weighted by how frequentlythe plant is cited. Thus, in order for a species to beranked highly, we must see not only consistency inuse reports but also a relatively high frequency ofreporting.For this study, each plant was considered indi-

vidually, even if it was used in combination withother species for a remedy. In addition, informationabout plant use received from household membersother than the key informant was used in the overalldata analysis, but excluded from analyses when thedata were subdivided by demographics.Once values were assigned to each plant species

using each index, the plants were assigned


importance ranks based on each index using theBrank.avg^ formula in Microsoft Excel. The specieswith the highest index score was given a rank of 1. Acombined list of the most highly ranked speciesacross the six indices was developed by adding therank values and selecting those species with thelowest totals (because plants with a rank of 1 aremost important).

We also calculated Spearman rank order correla-tions (for ranks with tied values) of the importanceranks corresponding to each index. The correlationswere calculated using SAS PROC CORR function.The resulting Spearman correlation coefficientsacross different importance rankings are all positiveand are reported as correlation matrices formattedwith darker cells for larger correlations and lightercells for smaller correlations.

For this analysis, the use categories in the datasetwere aggregated in three different ways. In the Balluses^ analysis, every illness use was considered as itwas recorded (n = 103). The Buses refined^ analysisgrouped reports into 37 illness categories, while theBuses category^ analysis grouped reports into 13 ill-ness categories (Appendix–Electronic SupplementalMaterial [ESM]). For example, sore on chest, boils,chickenpox, and skin rash were counted separately inthe Ball uses^ analysis, grouped into Bsores andrashes^ in the Buses refined^ analysis, and lumpedinto a dermatological category in the Buses category^analysis. When comparing responses by gender orgeneration, the Buses category^ aggregation was used.The total number of respondents, reports, and speciesanalyzed for each comparison is provided in Table 1.

Results and Discussion

In total, 2,262 reports of medicinal plant usewere collected and analyzed.

SUMMARY OF MOST IMPORTANT PLANTS

The 20 most highly-ranked species across the sixindices are shown in Table 2 (see Caldwell 2007 fora list of all species). The majority (61%) of thespecies on this list are trees, 22% are herbs, andthe remainder are woody shrubs and vines. Many ofthe most important species are not forest species(26%), but instead are found close to homes(48%) and/or cultivated (48%). In addition, non-native species appear equally as important as nativespecies (50/50%).

COMPARISON OF INDICES FOR ENTIRE DATA SET

Importance rankings of the entire dataset (notjust the top 20 species) produced by frequencycited, number of uses, cultural value, and relativeimportance indices are all highly correlated (r >0.9)(Table 3). That is, species ranked high by one indexare likely to be ranked high by another (displayed asa darker cell). This result is consistent with results ofother studies that compare similar quantificationindices (Albuquerque et al. 2006; Reyes-García etal. 2006). This result suggests that the index meth-od does not have a major impact on the importanceranking.

The informant consensus method (IAS) is theleast correlated with the other indices (generally, r<0.8). This means that a plant that is cited forhaving many uses (i.e., treating many illnesses) isnot likely to also be reported consistently for thesame uses by all respondents.

The new LIC method splits the difference be-tween the results for informant consensus and thefrequency cited, number of uses, cultural value, andrelative importance methods (generally 0.8 <r <0.9).This is consistent with expectations since the methodstarts with consensus and then weights it by frequen-cy cited. All of the correlations are positive andsignificantly different from zero at the <0.001 level.

INFLUENCE OF DATA SUMMARY: USE

AGGREGATION

In addition to high correlation between indicesconstructed using different formulae, there is alsohigh correlation between indices constructed usingdifferent use aggregations (Table 3). This includesthe index that is based on the number of uses, whichhas a r>0.95 correlation across different useaggregations.

Previous studies have suggested that the indexbased on reported number of uses can be biased bythe researcher’s subjective definition of use categories(Gausset 2004; Hoffman and Gallaher 2007). Ourresults indicate that this is not necessarily the case.

INFLUENCE OF SAMPLE SELECTION: RESPONDENT

DEMOGRAPHICS

Sample Disaggregation by Sex

As with the entire data set, all indices constructedfrom the responses by one sex (women or men)


result in similar, highly correlated rankings, regard-less of index used (Table 4). However, unlike theuse aggregation, sex does have a pronounced influ-ence on ranking results. Rankings provided by menare not highly correlated to responses by women(generally, r <0.5). The correlations between maleand female LIC indices were not all significant.There are no statistically significant differences

between males (N = 64 ) and females (N = 59) inthe average number of plants named or the averagenumber of illnesses named in an interview

(Caldwell 2007), but the species named, as indicat-ed by these correlation results, are very different.Because men and women differ in gender roles,their relationships with their environment, and theillnesses they experience, it makes sense that theyalso differ in their level and type of medicinal plantknowledge (Hanazaki et al. 2000; Kristensen andLykke 2003). In this study, womenmore frequentlyreport species found around the home and fuel-wood, while men report more marketed species,timber species, and farm species. Women are also

TABLE 1. TOTAL NUMBER OF RESPONDENTS, REPORTS, AND SPECIES USED IN ANALYSES (N VALUES).

Data Set # of Respondents # of Reports # of Species

Females 74 836 145Males 68 1,044 184Age 18–40 46 603 123Age 41–60 45 571 125Age 61+ 50 698 152Prompted Responses 112 710 138Unprompted Responses 115 1,037 169

TABLE 2.MEDICINAL PLANTS MOST HIGHLY RANKED BY IMPORTANCE INDICES.

Botanical Name** Local Name± Native Habit Habitat*

1 Khaya grandifoliola C.DC. mahogany/ koolu Yes Tree For, sav2 Alstonia boonei De Wild. tonton/ nyamedua Yes Tree For, sav, dist3 Ocimum gratissimum L. dzeveti Yes Herb surr, dist, cult4 Mangifera indica L. mango No Tree Sav, surr, dist, cult5 Vernonia amygdalina Delile gboti Yes Shrub Sav, surr, cult6 Citrus limon (L.) Osbeck lemon No Tree Surr, cult7 Species unidentified melonku Tree Sav8 Senna siamea (Lam.) H.S.Irwin & Barneby sangara No Tree Dist, surr, cult9 Azadirachta indica A.Juss. liliti No Tree Surr, cult10 Zingiber officinale Roscoe ginger/ancrasa No Herb Sav, cult11 Elaeis guineensis Jacq. palm/edeti Yes Tree Dist, for, cult12 Neonauclea excelsa (Blume) Merr. nymoke Yes Tree Distfor, sav, surr13 Sida acuta Burm.f. ademademe No Herb Surr, dist14 Combretum sp. Loefling aveto Yes Vine For15 Abrus precatorius L. dedekude No Vine Dist, distfor, surr16 Cola sp. Schott & Endl. agawu Tree17 Cymbopogon citratus (DC.) Stapf teagbe No Herb, grass Surr, cult18 Blighia sapida K.D.Koenig atiya Yes Tree For, sav, cult19 Newbouldia laevis (P.Beauv.) Seem. kpotiyia Yes Tree, shrub Distfor, sav, forest edge20 Xylopia aethiopica (Dunal) A.Rich. etso No Tree For, cult

*Habitat information from oral interviews. Sav = savanna; for = closed forest; distfor = disturbed forest/open woodland; dist =disturbed areas; surr = surroundings, e.g., gardens close to the home; cult = cultivated species, found in gardens/farms.**Camilian (herb, ranked #20) was excluded from this table and the analyses; neither the scientific name nor habitat could beidentified. The scientific name for melonku, ranked #7, could also not be identified. Nomenclature per The Plant List (2013).± The local names reported here are the names that were most commonly reported for the species during interviews. Most ofthese names are in the Ewe language, which is spoken in the area, but English common names were also often used for somespecies (mahogany, ginger, palm, and lemon). Nyamedua (rank #2) is in the Twi language.


more likely to report treatments for birth and deliv-ery, fever, stomach pain, neurological ailment, andworms (frequently affecting children), while menreport treatments for impotency, piles, cuts, andsnakebites (Caldwell 2007). These results are con-sistent with observations of gender roles, sex-specificailments, and location (i.e., where men and womenspend time).

Sample Disaggregation by Age

The results noted for sex were also seen when thedataset was analyzed by age cohorts. Within eachage class, rankings were highly correlated regardless

of the index used (r >0.9); however, correlation wasweaker (r <0.6) when comparing rankings across agegroups (Table 5). For this analysis, all of the corre-lations are significantly different from zero at the<0.001 level.

As with the analysis by sex, there was no statisticaldifference between age groups in the number ofplants or illnesses named (Caldwell 2007), but thedifferent age groups reported different species, asindicated by the correlation results. It is not surpris-ing that older people also cited more uses for a givenplant and a greater variety of medicinal uses, andtended to identify less commonly reported species.The middle-aged generation was more likely to cite

TABLE 3. CORRELATION MATRIX: INDEX COMPARISON BY USE AGGREGATION*.

All Uses Uses Refined Use CategoriesFreq #Uses CVs RI IAS LIC Freq #Uses CVs RI IAS LIC Freq # Uses CVs RI IAS LIC

All

Use

s

Freq 1 0.94 0.99 0.99 0.82 0.89 1 0.93 0.99 0.93 0.81 0.91 1 0.9 0.98 0.98 0.81 0.93

# Uses 1 0.98 0.98 0.61 0.76 0.94 0.99 0.97 0.99 0.63 0.79 0.94 0.96 0.97 0.97 0.65 0.84CVs 1 1 0.75 0.85 0.99 0.96 1 0.96 0.75 0.87 0.99 0.94 1 0.99 0.76 0.91

RI 1 0.74 0.84 0.99 0.96 1 0.96 0.75 0.87 0.99 0.94 0.99 0.99 0.76 0.91IAS 1 0.95 0.82 0.61 0.75 0.61 0.94 0.92 0.82 0.59 0.75 0.73 0.86 0.88LIC 1 0.89 0.76 0.85 0.76 0.89 0.98 0.89 0.74 0.85 0.83 0.82 0.95

Use

s R

efin

ed

Freq 1 0.93 0.99 0.93 0.81 0.91 1 0.9 0.98 0.98 0.81 0.93# Uses 1 0.97 1 0.57 0.76 0.92 0.98 0.97 0.97 0.6 0.81

CVs 1 0.97 0.74 0.87 0.99 0.94 1 0.99 0.76 0.90RI 1 0.57 0.76 0.92 0.98 0.97 0.97 0.6 0.81

IAS 1 0.93 0.81 0.57 0.74 0.73 0.92 0.89LIC 1.00 0.91 0.75 0.86 0.85 0.86 0.97

Use

Cat

egor

ies Freq 1 0.9 0.98 0.98 0.81 0.93

# Uses 1 0.96 0.96 0.53 0.77CVs 1 1 0.73 0.89

RI 1 0.72 0.88IAS 1 0.91LIC 1.00

*The higher the correlation, the darker the cell.

TABLE 4. CORRELATION MATRIX OF INDEX COMPARISONS BY SEX.

Female MaleFreq # uses CVs RI IAS LIC Freq # uses CVs RI IAS LIC

Fem

ale

Freq1 0.97 0.99 0.99 0.95 0.98 0.5 0.42 0.48 0.48 0.41 0.15

# uses 1 0.99 0.99 0.86 0.92 0.45 0.41 0.45 0.44 0.34 0.11*

CVs 1 1 0.92 0.96 0.49 0.42 0.47 0.47 0.39 0.14

RI 1 0.91 0.96 0.49 0.42 0.47 0.47 0.39 0.14IAS 1 0.98 0.43 0.33 0.4 0.4 0.35 0.10*LIC 1 0.48 0.39 0.45 0.45 0.38 0.12*

Male

Freq 1 0.94 0.99 0.99 0.89 0.7# uses 1 0.98 0.98 0.74 0.62CVs 1 1 0.84 0.66RI 1 0.84 0.66IAS 1 0.75LIC 1

*Only comparisons not significant at the <0.05 level. P values ranged from 0.06 to 0.15.


species found close to the home or purchased, per-haps reflecting their busy life stage and accumulatedwealth. The types of illnesses cited by the oldestcohort also differ from the youngest cohort, consis-tent with expectations about which illnesses affectpeople in different stages of life (Caldwell 2007).

INFLUENCE OF INTERVIEW STRUCTURE:PROMPTED VS. UNPROMPTED RESPONSES

To compare the prompted and unprompted re-sponses, we considered plants within the promptedillness categories only.The pattern observed for prompted vs. un-

prompted responses matches that observed for dataseparation by sex: intra-comparisons of indices are

highly correlated (r >0.9) while inter-comparisoncorrelation is weaker (r <0.5) (Table 6). For thisanalysis, all of the correlations are significantly dif-ferent from zero at the <0.001 level.The species cited most frequently without

prompting were: mahogany (Khaya grandifoliola),tonton (Alstonia boonei), melonku (species nameunknown), sangara (Senna siamea), and liliti(Azadirachta indica). All five of these are tree speciesthat are very commonly reported and used through-out the area. The five species cited most frequentlywhen prompted were: melonku (species name un-known), dzeveti (Ocimum gratissimum), coffee(Coffea canephora), ginger (Zingiber officinale), andmahogany (Khaya grandifoliola). Dzeveti and gingerare both herbs. Dzeveti is most frequently cited for

TABLE 5. CORRELATION MATRIX COMPARING INDEX ANALYSES BY AGE.

Ages 18–40 Ages 41–60 Ages 60+Freq #Uses CVs RI IAS LIC Freq #Uses CVs RI IAS LIC Freq #Uses CVs RI IAS LIC

Age

s 18

-40

Freq1.00 0.98 1.00 1.00 0.97 0.99 0.50 0.44 0.48 0.48 0.46 0.48 0.54 0.48 0.52 0.54 0.52 0.49

# Uses 1.00 0.99 0.99 0.93 0.96 0.47 0.42 0.45 0.45 0.42 0.45 0.52 0.47 0.50 0.52 0.50 0.46CVs 1.00 1.00 0.95 0.97 0.49 0.43 0.47 0.47 0.45 0.47 0.53 0.48 0.52 0.53 0.52 0.48RI 1.00 0.95 0.97 0.49 0.43 0.47 0.47 0.45 0.47 0.53 0.48 0.52 0.53 0.52 0.48IAS 1.00 0.99 0.47 0.41 0.45 0.45 0.42 0.44 0.48 0.43 0.46 0.48 0.46 0.43LIC 1.00 0.51 0.44 0.49 0.48 0.46 0.48 0.51 0.46 0.49 0.51 0.49 0.46

Age

s 40

-60 Freq 1.00 0.97 0.99 0.99 0.97 0.98 0.45 0.40 0.43 0.46 0.43 0.41

# Uses 1.00 0.99 0.99 0.91 0.94 0.40 0.35 0.38 0.40 0.38 0.36CVs 1.00 1.00 0.94 0.96 0.44 0.38 0.41 0.44 0.41 0.39RI 1.00 0.94 0.96 0.43 0.38 0.41 0.44 0.41 0.39IAS 1.00 0.99 0.40 0.35 0.38 0.40 0.38 0.35LIC 1.00 0.42 0.37 0.40 0.42 0.40 0.37

Age

s 60

+ Freq 1.00 0.97 0.99 1.00 0.99 0.95# Uses 1.00 0.99 0.97 0.99 0.89CVs 1.00 0.99 1.00 0.92RI 1.00 0.99 0.95IAS 1.00 0.92LIC 1.00

TABLE 6. CORRELATION MATRIX FOR INDEX RESULTS FOR PROMPTED VS. UNPROMPTED RESPONSES.

Ages 18–40 Ages 41–60 Ages 60+Freq #Uses CVs RI IAS LIC Freq #Uses CVs RI IAS LIC Freq #Uses CVs RI IAS LIC

Age

s 18

-40

Freq1.00 0.98 1.00 1.00 0.97 0.99 0.50 0.44 0.48 0.48 0.46 0.48 0.54 0.48 0.52 0.54 0.52 0.49

# Uses 1.00 0.99 0.99 0.93 0.96 0.47 0.42 0.45 0.45 0.42 0.45 0.52 0.47 0.50 0.52 0.50 0.46CVs 1.00 1.00 0.95 0.97 0.49 0.43 0.47 0.47 0.45 0.47 0.53 0.48 0.52 0.53 0.52 0.48RI 1.00 0.95 0.97 0.49 0.43 0.47 0.47 0.45 0.47 0.53 0.48 0.52 0.53 0.52 0.48IAS 1.00 0.99 0.47 0.41 0.45 0.45 0.42 0.44 0.48 0.43 0.46 0.48 0.46 0.43LIC 1.00 0.51 0.44 0.49 0.48 0.46 0.48 0.51 0.46 0.49 0.51 0.49 0.46

Age

s 40

-60 Freq 1.00 0.97 0.99 0.99 0.97 0.98 0.45 0.40 0.43 0.46 0.43 0.41

# Uses 1.00 0.99 0.99 0.91 0.94 0.40 0.35 0.38 0.40 0.38 0.36CVs 1.00 1.00 0.94 0.96 0.44 0.38 0.41 0.44 0.41 0.39RI 1.00 0.94 0.96 0.43 0.38 0.41 0.44 0.41 0.39IAS 1.00 0.99 0.40 0.35 0.38 0.40 0.38 0.35LIC 1.00 0.42 0.37 0.40 0.42 0.40 0.37

Age

s 60

+ Freq 1.00 0.97 0.99 1.00 0.99 0.95# Uses 1.00 0.99 0.97 0.99 0.89CVs 1.00 0.99 1.00 0.92RI 1.00 0.99 0.95IAS 1.00 0.92LIC 1.00


treatment of headaches while ginger is most com-monly cited for colds. Coffee is most commonlycited for the treatment of cuts.

The explanation for differences between speciescited in response to prompts vs. free-listing may liein how respondents think through free-listing exer-cises. When free-listing, respondents most oftenstarted with their most frequently used medicinalplant, inmany cases, mahogany. This species is usedto treat malaria, one of the most common illnessesin the community. Rather than naming additionalspecies after mahogany is mentioned, most respon-dents proceeded to list the numerous illnesses ma-hogany could be used to treat, such as body pain,stomach problems (e.g., pain, diarrhea, constipa-tion), anemia, headache, and fever. If one of theseven prompted illness categories was listed duringthis account, it was not prompted later. The respon-dents may have forgotten to list other plants if theywere not prompted by the illness. For example, ifheadache was the prompt, dzeveti might be listed asthe medicinal plant treatment; however, if headachewas mentioned under the mahogany free-listing, itwould not have been prompted and the respondentwould not have been triggered to think of dzeveti.

Conclusions

Ethnobotanical indices are tools employed tomeasure many types of information, e.g., medicinalplant diversity, frequencies with which plants areused as medicines, numbers of informants thatmention certain plants, or distribution of knowl-edge within a community. Depending on the focusof the research questions, certain methods will pro-vide more robust results. However, in our case studyof medicinal plant use in Ghana, we find that thespecific formula (or index) does not have muchinfluence on importance ranking results. All rank-ings are positively correlated with each other, withthe weakest correlation between the rankingresulting from the informant consensus methodand the rankings resulting from other methods.The new logarithmic informant consensus (LIC)method weights consensus values by the log ofhow frequently the species is cited. As expected,the results of the LIC method fall between thefrequency cited index and the informant consensusindex. When researchers are interested in both con-sensus and frequency cited, LIC offers an appropri-ate way to index plant importance. While this studyshows that six indices are highly correlated, each

index is different. Therefore, researchers should de-liberately assess their research questions in order toselect the most appropriate index for data analysis.

Like construction of the index, the level of aggre-gation of uses does not substantially affect the rank-ings. However, the species rankings are sensitive towho is interviewed and whether they are promptedby questions about specific illnesses. Thus, studiesthat sample specific demographics may find differ-ent species to be important compared to studies thatgather information from a representative sample ofthe entire community. This also requires delibera-tion in selecting a sample and an interview methodappropriate to the specific research question, sincethose initial choices are potentially more importantthan later choices about how to construct indicesand rank species.

The demographic results from our study suggestdirections for future research. We hypothesize thatmale involvement in labor practices farther from thehomestead has resulted in differential knowledgeabout particular medicinal plants that is not sharedby women. Similarly, middle-aged people also differmarkedly from younger and older age sets in theirmedicinal plant knowledge due to greater involve-ment in external labor markets. Future analyses willexamine the differences in plant knowledge betweencommunities and individuals exhibiting greater in-tegration into market economies.

AcknowledgmentsThis research was conducted in collaboration

with the Ghana Wildlife Society, including directorDr. Eramus Owusu, former site director Dr. EdemEkpe, site director Reuben Ottou, researcherRebecca Tettey, and chief Togbe Adabra. Localforest guides and herbalists were invaluable, servingas interpreters and providing expert knowledge ofthe local plants (David Logotse, Patrick Amexo,Amevor, Marcel, William, Johnson, and Issac). Aspecial thanks to the residents of the five surveyedvillages, who contributed their knowledge and ex-periences. Dr. Fikret Isik of North Carolina StateUniversity also provided assistance with the statisti-cal analyses presented in this report.

Literature CitedAbbiw, D. 1990. Useful plants of Ghana : West

African uses of wild and cultivated plants. ITPublications, Kew Royal Botanic Gardens,London.


Albuquerque, U. P., R. F. P. Lucena, J. M.Monteiro, A. T. N. Florentino, and C. B. R.Almedia. 2006. Evaluating two quantitative eth-nobotanical techniques. Ethnobotany Researchand Applications 4:51–60.

Amiguet, V. T., J. T. Arnason, et al. 2005. Aconsensus ethnobotany of the Q’eqchi’ Mayaof Southern Belize. Economic Botany 59(1):29–42.

Ayensu, E. 1978. Medicinal plants of West Africa.Reference Publications, Inc., Algonac, Michigan.

———, A. Adu, and E. Barnes. 1996. Ghana:Biodiversity and tropical forestry assessment.Prepared for USAID Contract P10/T641-0110-3-50035 by Edward S. Ayensu AssociatesLimited.

Begossi, A., N. Hanazaki, and J. Y. Tamashiro.2002. Medicinal plants in the Atlantic Forest(Brasil): Knowledge, use and conservation. Hu-man Ecology 30(3):281–299.

Bennett, B. C. and G. T. Prance. 2000. Introducedplants in the indigenous pharmacopeia of northernSouth America. Economic Botany 54:90–102.

Bonet, M. A. and J. Vallès. 2003. Pharmaceuticalethnobotany in the Montseny Biosphere Reserve(Catalonia, Iberian Peninsula): General results andnew or rarely reported medicinal plants. Journal ofPharmacy and Pharmacology 55(2):259–270.

Caldwell, K. I. 2007. Assigning medicinal plantvalue and estimating traditional environmentalknowledge in Ghana, Africa, using ethnobotan-ical measures. Master’s thesis, North CarolinaState University. http://repository.lib.ncsu.edu/ir/handle/1840.16/771.

Case, R. J., G. F. Pauli, and D. D. Soejarto. 2005.Factors in maintaining indigenous knowledgeamong ethnic communities of Manus Island.Economic Botany 59(4):356–365.

Gausset, Q. 2004. Ranking local tree needs andpriorities through an interdisciplinary action re-search approach. The Journal of Transdisciplin-ary Environmental Studies 3.

Gomez-Beloz, A. 2002. Plant use knowledge of theWinikina Warao: The case for questionnaires inethnobotany. Economic Botany 56(3):231–241.

Hanazaki, N., J. Y. Tamashiro, H. F. Leitao-Filho,and A. Begossi. 2000. Diversity of plant uses intwo Caiçara communities from the Atlantic For-est coast, Brazil. Biodiversity and Conservation9:597–615.

Hawthorne,W. 1990. Field guide to the forest treesof Ghana. Overseas Development Agency, Lon-don, U.K.

Heinrich, M., A. Ankli, et al. 1998. Medicinalplants in Mexico: Healers’ consensus and cultur-al importance. Social Science and Medicine47(11):1859–1871.

Hoffman, B. and T. Gallaher. 2007. Importanceindices in ethnobotany. Ethnobotany Researchand Application. 5:201–218.

Iwu, M. M. 1993. Handbook of African medicinalplants. CRC Press, Boca Raton, Florida.

Johns, T., J. O. Kokwaro, and E. K. Kimanani.1990. Herbal remedies of the Luo of Siaya Dis-trict, Kenya: Establishing quantitative criteria forconsensus. Economic Botany 44(3):369–381.

Kala, C. P., N. A. Farooquee, and U. Dhar. 2004.Prioritization of medicinal plants on the basis ofavailable knowledge, existing practices and usevalue status in Uttaranchal, India. Biodiversityand Conservation 13:453–469.

Kpordugbe, P. 2000. Small scale enterprise develop-ment Project Gbledi Traditional Area, Final report.Unpublished. Report for the Afadjato CommunityForest Conservation Project, January 2000.

Kremen, C., I. Raymond, and K. Lance. 1998. Aninterdisciplinary tool for monitoring conserva-tion impacts in Madagascar. Conservation Biol-ogy 12(3):549–563.

Kristensen, M. and A. M. Lykke. 2003. Informant-based valuation of use and conservation prefer-ences of Savanna trees in Burkina Faso. Eco-nomic Botany 57(2):203–217.

Kvist, L. P., M. Andersen, M. Hesselsøe, and J. K.Vanclay. 1995. Estimating use-values and rela-tive importance of Amazonian flood plain treesand forests to local inhabitants. CommonwealthForestry Review 74(4):293–300.

La Torre-Cuadros, M. A. and G. M. Islebe. 2003.Traditional ecological knowledge and use of veg-etation in southeastern Mexico: A case studyfrom Solferino, Quintana Roo. Biodiversityand Conservation 12:2455–2476.

Lykke, A. M., M. K. Kristensen, and S. Ganaba.2004. Valuation of local use and dynamics of 56woody species in the Sahel. Biodiversity andConservation 13:1961–1990.

Martinéz, G. J., A. M. Planchuelo, E. Fuentes, andM. Ojeda. 2006. A numeric index to establishconservation priorities for medicinal plants inthe Paravachasca Valley, Córdoba, Argentina.Biodiversity and Conservation 15:2457–2475.

Owusu E. H. 2001. Community-based conserva-tion in Ghana: The potential of the Afadjato andAgumatsa Range for ecotourism. Ph.D. thesis,University of Kent at Canterbury, Kent, U.K.


http://repository.lib.ncsu.edu/ir/handle/1840.16/771

http://repository.lib.ncsu.edu/ir/handle/1840.16/771

Pardo-de-Santayana, M. 2003. Las plantas en lacultura tradicional de la Antigua Merindad deCampoo. Ph.D. thesis, Departamento deBiología, Facultad de Ciencias, UniversidadAutónoma de Madrid, Spain

Phillips, O. and A. H. Gentry. 1993. The usefulplants of Tambopata, Peru: I. Statistical hypoth-eses tests with a new quantitative technique.Economic Botany 47:15–32.

——— and ———. 1993b. The useful plants ofTambopata, Peru: II. Additional hypothesis test-ing in quantitative ethnobotany. Economic Bot-any 47:33–43.

———, ———, C. Reynel, P. Wilkin, and C.Galvez-Durand. 1994. Quantitative ethnobota-ny and Amazonian conservation. ConservationBiology 8:255–248.

The Plant List. 2013. www.theplantlist.org (July2015).

Prance, G. T., W. Balée, B. M. Boom, and R. L.Carneiro. 1987. Quantitative ethnobotany andthe case for conservation in Amazonia. Conser-vation Biology 1(4):296–310.

Reyes-García, V., V. Vadez, E. Byron, L. Apaza,W.R. Leonard, E. Perez, and D. Wilkie. 2005.Market economy and the loss of folk knowledgeof plant uses: Estimates from the Tsimane’ of theBolivian Amazon. Current Anthropology 46:651–656.

———, T. Huanca, V. Vadez, W. Leonard, and D.Wilkie. 2006. Cultural, practical, and economic

value of wild plants: A quantitative study in theBolivian Amazon. Economic Botany 60(1):62–74.

Rossato, S. C., H. Leitao Filho, and A. Begossi. 1999.Ethnobotany of caiçaras of the Atlantic ForestCoast (Brazil). Economic Botany 53:387–395.

Silva, V. A., L. H. C. Andrade, and U. P. Albu-querque. 2006. Revising the cultural significanceindex: The case of the Fulni-ô in northeasternBrazil. Field Methods 18:98–108.

Swaine, M. D., P. Ekpe, F. Seku, and M. Abu-Juam. 1999. Vegetation survey of the Mt.Afadjato Community Forest Conservation Pro-ject. Unpublished. Report prepared for the Gha-na Wildlife Society, December 1999.

Tardío, J. andM. Pardo-de-Santayana. 2008. Culturalimportance indices: A comparative analysis basedon the useful wild plants of Southern Cantabria(Northern Spain). Economic Botany 62(1):24–39.

Tettey, R., E. Ekpe, and E. Owusu. 2005. Unpub-lished data collected on non-timber forest prod-ucts in the Afadjato area.

Trotter, R. T. and M. H. Logan. 1986. Informantconsensus: A new approach for identifying po-tentially effective medicinal plants. In: Plants inindigenous medicine and diet, ed. N. L. Etkin,91–112. Bedford Hill, New York: RedgravePublishing Company.

Voeks, R. A. and A. Leony. 2004. Forgetting theforest: Assessing medicinal plant erosion in East-ern Brazil. Economic Botany 58(Supplement):S284–S306.


http://www.theplantlist.org/

how study design influences the ranking of medicinal plant...

Documents