Identifying Key Vocabulary in Ecology Textbooks for ESP: A Corpus-Based Keyword Analysis

Abstract

On one hand, the University of Kragujevac, Serbia, needs an ESP course in ecology; on the other hand, the available wordlists do not meet fully meet the needs of the target group of students, i.e., the undergraduate ecology students. This discrepancy served the a main incentive for creating an institution-specific word list of key content words found in a corpus compiled from English-medium textbooks intended for ecology students (here referred to as the Eco-Text corpus). The keyword list was generated in AntConc, with a keyness measure threshold automatically set at p < 0.055 (3.84 with Bonferroni correction). After the list was complete, the NGSL (New General Service List), NAWL (New Academic Word List), and the SWL (Science Word List) Highlighters were used to further systemize the words into four sublists. In addition to presenting some key findings, we also suggest areas for further refinement. Although preliminary, the findings can be useful to practitioners in planning an ESP course in ecology and developing materials.

1. Introduction

Internationalization has become a central objective for many higher education institutions (HEIs) worldwide (Hartle et al., 2025). Since English is a lingua franca in academic domains, English—Medium of Instruction (EMI) has emerged as one of the dominant strategies to internationalize a curriculum and prepare students for international mobility, joint degrees, and a globalized workforce (Macaro et al., 2018; Galloway & Rose, 2021; Hartle et al., 2025). Serbia is not an exception. Most faculties at the University of Kragujevac, Serbia, a HEI where the authors of this paper are employed, have decided to introduce ESP courses to prepare students for EMI and internationalization in general.

An ESP course targeting ecology has yet to be developed for undergraduate students. The starting point for the course planning and materials design is to detect the vocabulary ecology students need. Vocabulary is central to understanding and producing language at any level, especially in ESP contexts (Hirsh & Coxhead, 2009). Specialized vocabulary is the core of any efficient ESP course, as it is the lexis that distinguishes various fields of expertise and even their specific domains (Mockiené et al., 2022).

This is where corpus linguistics enters the scene as “one of the great tools to answer the vocabulary needs of our ESP students” (Rayson & Potts, 2021). The current attempts to target the vocabulary salient to ecology do not meet the needs of the target student population. For instance, Liu and Han (2015) offer the Environmental Academic Word List (EAWL), but their purpose-built corpus comprises research articles. Wild et al. (2013) quantify lexical usage of the vocabulary pertaining to the environment and ecosystems in three custom-built specialized web corpora and a reference web corpus. Our target learners rely more heavily on textbooks. Hence, the vocabulary used in textbooks should be prioritized with the target group of learners.

The main goal of this paper is to target the words salient in ecology textbooks, i.e., to create the institution-specific Ecology Textbook Word List (ETWL). To do so, we compiled a purpose-built corpus comprising five English-medium ecology textbooks. A corpus-based keyword analysis was conducted with AntConc. The maximum number of hits in AntConc equals 5000 hits, and to minimize the possibility that key vocabulary is omitted due to the large number of high-frequency functional words and general vocabulary. Moreover, it is generally expected that upon completing secondary school, a large proportion of incoming Serbian freshmen have a B level of English language proficiency (ELP) (Mićović et al., 2023). In other words, freshmen are expected to already have a significant repertoire of functional words. Keyword analysis identifies domain-specific vocabulary while filtering out general high-frequency words.

Some studies have reported that Serbian freshmen are an inhomogeneous group in terms of their ELP (Danilović Jeremić, 2015; Ivančević-Otanjac, 2021; Mićović et al., 2023). Hence, we find it useful to include cross-sectional analyses of the ETWL and the NGSL (New General Service List), the NAWL (New Academic Word List), and the SWL (Science Word List) in order to create sublists that can be utilized as tailored support in accommodating students with varying language skills. The research questions defined here are as follows: RQ1: How many lemmas from the Eco-text keyword list are listed in the NGSL? RQ2: How many lemmas from the Eco-text keyword list are also listed in the NAWL? RQ3: How many lemmas from the Eco-text keyword list are also listed in the SWL? RQ4: How many lemmas from the Eco-text keyword list are not listed in the NGSL, NAWL, and the SWL? The lemmas not included in these lists are used to form the Ecology Textbook Word List (ETWL).

2. Literature Review

Vocabulary acquisition is the foundation of language learning (Sheng & McGregor, 2010). Positive correlations have been determined for vocabulary size and overall language proficiency (Schmitt, 2010), but also students’ writing skills (Hinkel, 2011; Viera, 2017; Kiliç, 2019), reading skills (Van Zeeland & Schmitt, 2012; Yuliastuti, 2021; Yan & Li, 2021; Lisana & Subroto, 2022; Iffah & Subroto, 2023; Pravitasari & Nurfadillah, 2025) and to a lesser extent, listening comprehension (Staehr, 2008).

The challenge of targeting key vocabulary for an ESP course has been facilitated with outstanding advances in software and computer technology over the past decades, since tremendous digital textual databases can be analyzed with relative ease. The research focusing on targeting discipline-specific vocabulary and its frequency has been intensified with the rapid development of computer-assisted corpus analysis. Summarizing the numerous definitions present in the previous literature (cf. Sinclair, 1996; D. Biber et al., 1998; Müller & Waibel, 2000), McEnery et al. (2006) conclude that a corpus is a collection of machine-readable, authentic texts that is sampled to be representative of a particular language, language variety, or register.

The automaticity of the process enabled by advancements in computer technology has resulted in the proliferation of word lists. Several high frequency lists have been compiled, such as GSL (General Service List) by West (1953), the NGSL 1 (New GSL1) by Browne (2014), and the NGSL 2 (New GSL2) by Brezina and Gablasova (2015). The importance of Academic vocabulary has also been underscored in corpus-based word lists. It should be noted that academic vocabulary differs from both the general high-frequency words and discipline-specific vocabulary in that it includes “formal, context-independent words with a high frequency and/or wide range of occurrences across scientific disciplines” (Farrell, 1990). Three major word lists of general academic vocabulary have been developed since 2000: AWL (Academic Word List) by Coxhead (2000), NAWL (New AWL) by Browne et al. (2013) and the AVL (Academic Vocabulary List) by Gardner and Davies (2014). Field-specific academic word lists (Martínez et al., 2009; Liu & Han, 2015) or discipline-based lexical repertoires (Hyland & Tse, 2007) focus on the academic vocabulary closely related to specific disciplines, such as pharmacy (Heidari et al., 2020), accounting (Khany & Kalantari, 2021), linguistics (Moini & Islamizadeh, 2016), geography (Drašler & Kavalir, 2026), marine engineering (Đurović, 2021), etc. Discipline-specific vocabulary typically has low frequency in both general language and across academic disciplines in general (Qureshi & Malik, 2025). It contains vocabulary that is specific to a discipline and related fields, such as electronics (e.g., Farrell, 1990), engineering (e.g., Ward, 1999), medicine (e.g., Wang et al., 2000), business (e.g., Browne & Culligan, 2015), or even plumbing (Coxhead & Demecheleer, 2018), and baseball (Riccobono, 2020). The interest in specific sub-fields stems from the fact that each discipline has its own conventions (Yang, 2015), so available general academic word lists commonly fail to provide sufficient lexical coverage of specialized words (G. Biber, 2006; Durrant, 2016).

Tong et al. (2025) provide a systematic review of 54 corpus-based wordlists developed between 2000 and 2024. In 50 articles, frequency is used as one of the selection criteria. In most studies, at least one more selection criterion is employed, including specialized occurrence (typically ensured by excluding general words), dispersion (also called distribution), keyness, discipline (excluding lemmas concentrated in specific parts of the corpus), expert judgement, and/or dictionary checking. Frequency refers to the total occurrences of a lexical item in a corpus. High-frequency words are commonly seen as the vital starting point for learning vocabulary (Nation, 2013; Schmitt, 2010) since lexemes that occur most frequently are those that each learner is more likely to meet or need (Sutarsyah et al., 1994; Nation, 2006; Granger, 2015; Riccobono, 2020).

As reported by Tong et al. (2025), keyness is used in only seven articles. A keyword is commonly defined as a word occurring with unusual frequency in a text (Scott, 1997), identified by “statistical comparison” of a target corpus with reference or benchmark corpora, which most commonly represent general language (Evison, 2010). Differences in frequency values are compared using the log-likelihood test, which determines whether the observed differences in frequencies are statistically significant or just chance occurrences (Szudarski, 2017). In other words, the software automatically counts the frequencies in a purpose-built and a benchmark corpus, calculates the log-likelihoods, and generates the keyword list, i.e., the list of words that are more frequent in a specialized corpus as compared to one chosen to represent general language. The fact that keyness is used as a selection criterion in only seven articles analyzed in Tong et al. (2025) does not imply nor suggest that a keyness analysis is by any means deficient. Quite contrary, to avoid the acquisition of a large vocabulary stock that is of no use to learners while omitting technical words needed to understand language and communicate within educational and/or professional settings, ESP courses should focus on specialized vocabulary, i.e., words “specific to particular topic, field or discipline” (Nation, 2001). Hence, the analysis that compares the frequencies of words in a specialized corpus to a general corpus may have many advantages. It filters general vocabulary and brings focus to the lexical items that are salient in specific disciplines. It is important to note that in some of the studies analyzed by Tong et al. (2025), there was no need for keyness analysis because, for instance, they focused on general academic language (Coxhead, 2000; Coxhead & Hirsh, 2007).

3. Methodology

The goal of this paper is to target the key words in ecology textbooks as the first step toward the development of an ESP course for undergraduate ecology freshmen. Drawing on the methodology familiar to computer corpus linguistics (Anthony, 2018; Mockiené et al., 2022), the procedure includes four steps: (1) sampling (compiling a corpus consisting of five ecology textbooks published in English), (2) conversion of the .pdf files into a machine-readable .txt format (AntFileConverter 2.1.0), (3) Part-Of-Speech (POS) tagging (TagAnt 2.1.1), and (4) generating a keyword list (AntConc 4.3.1).

3.1. Sampling

The corpus analysis was conducted on an institution-specific, purpose-built, and specialized corpus of five ecology textbooks used by the target student population in the ecology courses: A Textbook of Ecology (Ahad & Ferdous, 2019), Ecology: from Individuals to Ecosystems (Begon & Townsend, 2006), Introduction to Ecology (Ford, 2013), Ecology and Applied Environmental Science (Hadjibiras, 2014), and Fundamentals of Ecology (Odum, 1971). A PDF editor was used to remove filter out the redundant pages, such as glossaries, reference lists, etc., and retain only the chapters, i.e., the pages with educational content. Files were converted to a .doc format to manually remove the hyphenation at line breaks using the “Find” option. The final corpus database (Eco-Text) comprises 509,890 tokens.

3.2. File Conversion and POS Tagging

The second and third phases refer to the preparation of the textual database for further computer processing. First, the textbooks must be converted from the .pdf files into a machine-readable format, i.e., .txt format. The software solution AntFileConverter 2.1.0 was used to create a textual database that other software solutions used here can operate on. The files were used to conduct a keyword analysis. The next step is to perform POS tagging. TagAnt 2.1.1, a freeware segmenter and Part-Of-Speech (POS) tagger, is utilized to obtain POS-tagged text material. This step guarantees that all words are tagged based on their part of speech. The POS-tagged files were further used to determine the part of speech in all those cases in which a word form can be used as a noun or a verb (e.g., breed, cause, etc.) and a noun or an adjective (e.g., alternative, relative, etc.).

3.3. Lexical Profiling

Finally, AntConc 4.3.1, a freeware corpus toolkit for concordancing and text analysis, was selected for computer-assisted lexical analysis. The operation keyword was selected to generate the keyword list. The program generates keyword lists automatically with a keyness measure threshold set at p < 0.055 (3.84 with Bonferroni correction). It compares the frequencies of words in a target corpus and a reference/benchmark corpus (AmE06) based on the background statistical analysis relying on log-likelihood values. The number of hits was set to the maximum possible value for this software solution, i.e., 5000 keywords. The target corpus is the purpose-built corpus of ecology textbooks, and the AmE06, incorporated into AntConc 4.3.1, is used as a reference (benchmark) corpus. The BrE06 database is also incorporated in AntConc but it has not been selected as a reference corpus because it has been decided that there was a risk that at least a certain set of words would be recognized by the software as keyword based solely on spelling. The BrE06 can be used in the future as a sort of a validation test. In addition, since our corpus is relatively small, we did not see a need to include any larger external database for an analysis that is still in its preliminary phase.

The list obtained had to undergo manual examination. First, additional limitations were set in terms of number of hits and range (no. of hits > 10 and range > 2). This means that only the words used more than 10 times in Eco-Text and in more than three textbooks are included in the keyword list. The POS-tagged files were used to determine the parts of speech and classify different word forms as lemmas. Lemmas are used to systematize the word forms rather than word families since not all family members are listed in the NGSL (New General Service List), NAWL (New Academic Word List), and the SWL (Science Word List). Namely, some frequent members are included in the NGSL while others may be found in the NAWL.

The NGSL, NAWL, and SWL Highlighters available at the EAP Foundation website are used to perform the cross-sectional analyses with the NGSL, NAWL, SWL, and the one obtained through this study. The latter list is entered into the box given on the NGSL page, and the software highlights the words that are included in the NGSL. The same principle is used with the other two highlighters. The vocabulary not detected in these three lists entered the ETWL, i.e., the word list containing lemmas salient in the selected ecology textbooks.

4. Results

The keywords detected through this analysis are grouped into lemmas and further classified in the NGSL, NAWL, SWL, and ETWL vocabulary. Table 1 provides an overview of parts of speech across these categories.

Table 1. Parts of speech across four sublists.

	NGSL	NAWL	SWL	ETWL	Total
Nouns	264	97	7	160	528
Verbs	110	20	3	14	147
Adjectives	89	43	16	72	220
Adverbs	19	5	5	14	43
Total	482	165	59	260

The keyword list obtained through the corpus analysis after adding the mentioned parameters (no. of hits > 10 and range > 2) and organizing word forms into lemmas has 929 items. Their distribution is shown in Figure 1.

Figure 1. Distribution of lemmas across four sublists.

4.1. The NGSL Sublist

The general vocabulary sublist is obtained by comparing the list obtained through this study with the NGSL (Browne et al., 2013). About 49% of key words from Eco-Text corpus are also listed in the revised general service list. Table 2 presents the top 40 lemmas in order of their absolute frequency.

Table 2. The top 40 lemmas in the NGSL.

Rank	Type	POS	Hits	Range	Rank	Type	POS	Hits	Range
1	species	n	4029	5	21	individual	n, adj	911	4
2	population	n	2716	3	22	resource	n	910	5
3	plant	n	2267	5	23	model	n, v	884	3
4	figure	n	1855	4	24	example	n	877	5
5	increase	n, v	1367	5	25	host	n	837	4
6	density	n	1312	5	26	change	n	830	5
7	community	n	1265	5	27	chapter	n	828	4
8	high	adj	1188	5	28	size	n	818	3
9	rate	n	1145	4	29	level	n	795	4
10	effect	n	1107	5	30	animal	n	790	5
11	number	n	1095	4	31	natural	adj	772	5
12	water	n	1068	5	32	different	adj	730	5
13	flood	n	1000	4	33	large	adj	725	5
14	environment	n	987	3	34	forest	n	708	5
15	most	adv	970	5	35	environmental	adj	697	5
16	case	n	964	4	36	small	adj	684	5
17	low	adj	941	4	37	temperature	n	684	4
18	many	adj	940	5	38	cycle	n, v	653	4
19	competition	n	938	4	39	energy	n	643	5
20	area	n	925	5	40	factor	n	634	4

It is important to note that the high share of the core vocabulary justifies our decision to focus on keywords rather than high-frequency words. The latter scenario would result in a higher share of general vocabulary at the cost of words salient in ecology textbooks. It can also be suggested that the inclusion of keywords with a range value equal to 2 would result in a lower share of general vocabulary and a higher share of the vocabulary that is salient in ecology textbooks.

4.2. The NAWL Sublist

The sublist of academic vocabulary is obtained by comparing the list obtained through this study with the NGSL (Browne et al., 2013). About 17% of key words from Eco-Text corpus are general academic vocabulary. Table 3 presents the top 40 lemmas in order of their absolute frequency.

Table 3. The top 40 lemmas in the NAWL.

Rank	Type	POS	Hits	Range	Rank	Type	POS	Hits	Range
1	ecology	n	1160	5	21	equilibrium	n	212	4
2	organism	n	1059	5	22	evolutionary	adj	195	4
3	predator	n	990	5	23	biology	n	194	4
4	prey	n, v	919	5	24	reproduction	n	190	4
5	habitat	n	673	5	25	tropical	adj	188	4
6	insect	n	620	4	26	dependence	n	187	3
7	ecological	adj	585	5	27	spatial	adj	169	4
8	nutrient	n	574	5	28	intensity	n	168	3
9	nitrogen	n	548	4	29	harvest	n, v	164	5
10	leaf	n	486	4	30	oxygen	n	159	3
11	dynamics	n	471	4	31	optimal	adj	156	3
12	productivity	n	462	4	32	biodiversity	n	155	4
13	distribution	n	441	5	33	offspring	n	153	4
14	niche	n	417	4	34	toxic	adj	146	3
15	radiation	n	306	3	35	consumption	n	143	4
16	mortality	n	285	4	36	availability	n	133	4
17	conservation	n	259	4	37	disturbance	n	133	4
18	bacteria	n	246	5	38	differentiation	n	125	4
19	vegetation	n	239	4	39	acid	n	119	4
20	fungus	n	213	5	40	oscillation	n	113	3

The NAWL comprises 570 word families with 3112 words and is designed to supplement the NGSL. From this perspective, it is reasonable to highlight that the NAWL and NGSL vocabulary constitutes 66% of the Eco-text keywords. In other words, even when combined, the mentioned lists provide less than half of the vocabulary stock ecology students need to comprehend their textbooks.

4.3. The SWL Sublist

About 6% of keywords in the Eco-text are also present in the SWL (Coxhead & Hirsh, 2007). The SWL, developed by Averil Coxhead and David Hirsh in 2007, contains 318 families not occurring in the GSL or AWL. With the NGSL and NAWL, the SWL nouns constitute 72% of the lemmas recorded as the key vocabulary in the selected ecology textbooks. The remaining nominal lexemes can be labelled as ETWL vocabulary. The top 40 words are presented in the Table 4.

Table 4. The top 40 lemmas in the SWL.

Rank	Type	POS	Hits	Range	Rank	Type	POS	Hits	Range
1	abundance	n	555	4	21	nitrate	n	75	3
2	parasite	n	518	4	22	recycle	v	73	3
3	dispersal	n	323	3	23	extinct	adj	67	4
4	terrestrial	adj	290	4	24	humidity	n	63	4
5	extinction	n	266	5	25	pollen	n	57	3
6	aquatic	adj	260	4	26	latitude	n	55	3
7	marine	adj	208	4	27	decompose	v	51	4
8	decomposer	n	198	4	28	disperse	v	51	3
9	photosynthesis	n	165	5	29	algal	adj	46	3
10	algae	n	161	4	30	ammonium	n	44	3
11	decomposition	n	158	4	31	metabolic	adj	44	4
12	temperate	adj	134	4	32	dispersion	n	39	4
13	atmospheric	adj	122	3	33	evaporation	n	39	3
14	respiration	n	117	4	34	spatially	adv	36	3
15	inorganic	adj	104	4	35	flora	n	31	4
16	climatic	adj	91	4	36	oxide	n	31	3
17	photosynthetic	adj	88	4	37	fertilizer	n	30	4
18	microbial	adj	83	3	38	hemisphere	n	30	3
19	abundant	adj	81	5	39	moisture	n	30	3
20	degradation	n	78	3	40	microbe	n	28	4

4.4. The ETWC Sublist

About 28% of keywords in the Eco-text are not present in the NGSL, NAWL, and SWL. Here they are labeled as ETWL (Ecology Textbook Word List) vocabulary. The data on the top 60 ETWL lemmas with their absolute frequency (AF), range (R), log likelihood (LL), and dispersion (D) are given in Table 5.

Table 5. The top 60 ETWL lemmas.

Type	POS	AF	R	LL	D	Type	POS	AF	R	LL	D
ecosystem	n	1111	3	2226.381	0.546	abiotic	adj	92	4	184.318	0.701
richness	n	544	3	1090.146	0.493	herbivory	n	91	3	182.315	0.615
predation	n	323	4	647.198	0.709	biotic	adj	89	4	178.308	0.726
biomass	n	309	4	619.141	0.703	polymorphism	n	88	3	176.304	0.612
trophic	adj	304	4	609.121	0.573	exploitation	n	87	4	76.198	0.728
herbivore	n	261	4	138.237	0.716	restoration	n	87	4	99.982	0.498
reproductive	adj	260	4	428.314	0.409	biosphere	n	85	3	170.294	0.406
global	adj	243	3	82.888	0.751	litter	n	85	3	170.294	0.638
ant	n	226	5	230.401	0.626	ozone	n	85	3	170.294	0.107
succession	n	204	4	408.731	0.566	parasitism	n	85	3	170.294	0.723
fecundity	n	179	3	358.636	0.542	removal	n	85	4	170.294	0.675
phosphorus	n	171	3	341.606	0.734	interference	n	84	3	168.29	0.638
graze	v	170	4	410.722	0.592	worm	n	84	4	168.286	0.694
ecologist	n	153	4	40.068	0.636	endanger	v	83	4	166.286	0.555
coastal	adj	152	3	304.536	0.586	microorganism	n	82	5	164.283	0.491
grassland	n	148	5	296.51	0.649	fauna	n	81	4	162.279	0.623
coexistence	n	147	4	294.517	0.569	exclusion	n	80	4	74.048	0.578
isocline	n	144	3	288.496	0.529	mutualistic	n	80	3	212.364	0.386
mammal	n	138	5	356.62	0.669	persistence	n	80	3	100.012	0.723
beetle	n	134	4	508.889	0.77	squirrel	n	80	4	160.273	0.652
invertebrate	n	123	3	246.422	0.689	reef	n	79	3	106.181	0.688
pollutant	n	116	4	232.399	0.487	volterra	n	79	4	158.272	0.487
freshwater	n	111	3	222.386	0.755	populate	v	78	3	240.412	0.491
coexist	v	109	3	130.223	0.476	detritus	n	76	3	152.262	0.506
mutualism	n	102	3	370.64	0.417	rodent	n	76	3	152.259	0.519
seasonal	adj	102	4	204.354	0.834	marsh	n	74	3	148.252	0.618
colonization	n	101	3	202.351	0.744	canopy	n	72	3	89.382	0.664
naturalist	n	95	4	190.329	0.06	pine	n	68	3	64.648	0.685
carnivore	n	94	4	188.323	0.728	intrinsic	adj	67	4	77.908	0.716
greenhouse	n	93	3	119.324	0.374	oak	n	66	3	49.81	0.729

5. Discussion

The software-generated word frequency lists are used in making decisions on whether lexical items should be taught or dismissed. In addition, corpus findings ensure that the limited classroom teaching time is devoted to those lexical items that “provide a good return for the learning effort” (Szudarski, 2017). This analysis targets the key vocabulary in the selected ecology textbooks (Eco-Text). Four sublists are created: the NGSL, NAWL, SWL, and ETWL. In many studies, specialized occurrence is ensured by eliminating general and/or general academic vocabulary (Liu & Han, 2015; Kim & Lee, 2019; O’Flynn, 2019; Jablonkai, 2020; Kim & Lee, 2023; Barraza et al., 2024). In this analysis, the NAWL and NGSL words are not eliminated from the list. They are preserved as sublists so that the practitioners who find this list useful can make unique combinations based on the needs of their students. If after prioritizing the ETWL, there is not enough time to underscore the items from the other sublists, students can be given lists or glossaries to learn at home. Pribabadi and Rahmany (2014) examined the effect of keyword method and word-list method instruction on ESP vocabulary learning and concluded that their results are consistent with the findings of previous studies showing that the former method is more effective.

The list of the most frequently occurring vocabulary in a specialized corpus is a valuable tool in designing an ESP syllabus in that it can be utilized to determine the lexical content. Vocabulary lists should be used to plan course syllabi, develop materials and textbooks, and prepare tests (Flowerdew, 1993; Jones & Durrant, 2010). The list provided in this paper can be used as a stand-alone list or can complement other materials. Word lists can be used as a foundation for developing glossaries or even dictionaries. More importantly, they can serve as a guide in developing ESP textbooks. Textbook content is commonly seen as one of the main factors contributing to enhanced student motivation and better language acquisition (Stojanović, 2014).

Although academic word lists for specialized areas are beneficial to optimizing vocabulary instruction and the overall learning outcomes (Ward, 2009), individual pedagogical implications and recommendations can serve only as a guideline; course design is to be specifically tailored to address the professional language needs in terms of other linguistic assets, language skills, and communicative competence. Hence, the language of ecology deserves more comprehensive analyses of the dominant word patterns, grammatical structures, and other discursive elements.

In addition, vocabulary instruction should be integrated with other language skills. On the one hand, vocabulary mastery has a positive impact on all skills. On the other hand, the development of language skills enables spaced repetition and contextualized practice, which are essential for long-term retention and reinforcing vocabulary learning (Nation, 2013). In other words, vocabulary should be learned and memorized via reading, listening, writing, and speaking. The words contained in these sublists should be introduced through reading and listening to authentic language material. By doing so, one can observe and analyze how words are used in different sentential contexts. In addition, ESP students should be encouraged to use the words through both productive skills: writing and speaking. The products of productive skills reveal whether learners know the precise meanings of words and how to use them in a sentence.

The findings of this analysis also indicate that further analyses are needed. Keyword analysis with the BrE06 as a reference corpus, or any other larger general language database, should be used to validate the findings of this study. In addition, word polysemy and homonymy should be investigated in detail before deciding to disregard any of the sublists. Even at first glance, it is easy to see that many NGSL words are polysemic. It is possible that one meaning of a word is used frequently in general language, while any of its other meanings is more prominent in ecology. For instance, the word root may be more frequent in general language, meaning the source or the origin of something, as in the phrase the root of the problem. The same word stands for a morpheme that underlies an inflectional or derivational paradigm (in grammatical and/or linguistic literature), a quantity that when raised to the nth power (multiplied by itself n times) produces a given quantity (in mathematics), and, inter alia, a part of the body of a plant (in biology, ecology, etc.).

The wordlists also show that morphological patterns deserve more attention in future studies. For instance, the verbal prefix de- to signify reversal (i.e., the opposite action) in decomposition (from decompose) and removal (i.e., reduction) in defoliation (from defoliate) and deforestation (from deforest) appears to be more dominant than is the case in general language. In addition, words from the four lists can be classified into word families. The benefits of learning vocabulary via derivational families include, inter alia, the multiplication of vocabulary size and improved reading comprehension, i.e., the recognition of word families helps learners guess the meaning of unfamiliar words based on their knowledge of root and affixes (Bauer & Nation, 1993; Carlisle, 2000; Nation, 2001; Kieffer & Lesaux, 2007).

Finally, the overlap between the SWL and the ETWL deserves more attention. The percentage of only 6% may be surprising without knowing the structure of the corpus used to create the SWL. The corpus of 1.7 million words covers 14 subject areas: agriculture, biology, chemistry, computer science, ecology, engineering and technology, geography, geology, horticultural science, mathematics, nursing and midwifery, physics, sport and health sciences, and veterinary and animal sciences.

6. Conclusions

The findings of the corpus-based keyword analysis of the selected ecology textbooks clearly indicate the necessity for discipline-specific ecology wordlists. The NGSL, NAWL, and SWL cannot provide sufficient vocabulary stock for undergraduate ecology students. On the other hand, courses based on these lists would provide students with a large vocabulary stock that is not immediately needed. Consequently, the main pedagogical implication of the findings presented here is that an ESP course in ecology is needed and worth the investment of time and resources.

The findings also underscore the importance of conducting a thorough needs analysis, including from different perspectives. Different target groups of students have different ELP levels and would need different combinations of the sublists defined in this paper. In addition, it is important to know the genre that students mostly rely on. For instance, the undergraduate freshmen in our context rely solely on textbooks, hence the need to conduct a keyword analysis on textbooks, without including other genres. Liu and Han’s (2015) word list of environmental science overlaps with the AWL by about 50%. The overlap with the academic vocabulary list used in this paper (NAWL) and the keywords in the Eco-text is significantly lower.

The vocabulary items detected through this study are presented in order of frequency. This does not necessarily mean that high frequency is the only indicator that certain lexical items should be prioritized. High frequency raises a chance that a student will encounter a word sooner than other, less frequent words. However, that does not have to be the case. In some contexts, range may be more relevant. Probably the best strategy is to establish cooperation with colleagues teaching the courses in the field. They can provide the chronological lists of the textbooks and/or topics. Hence, ESP practitioners can introduce vocabulary based on the order it is expected to be introduced in their subject area.

This analysis is just a preliminary study into the keywords in ecology textbooks, and should be expanded and deepened in the future. In addition to further frequency and keyword analyses and analyses of morphological and collocational patterns, future studies should also focus on grammatical patterns. Different discursive practices should also be explored. The main limitation is that even practitioners whose target learners also rely on ecology textbooks should use these word lists with caution. Ideally, all ESP practitioners should be equipped with the corpus analysis skills to create institutionally-specific corpora and word lists. Only with such material can one be sure that the corpus is absolutely representative of the English-medium teaching material students have to use in their fields.