Vocabulary is the most fundamental aspect in language, and the importance of vocabulary has been constantly repeated (Nation, 2013; Webb, 2020). Together with the development of wordlists, such as the British National Corpus (BNC) lists (Nation, 2006) and the British National Corpus/Corpus of Contemporary American English (BNC/COCA) lists (Nation, 2017), researchers in the field of vocabulary studies have been continuously giving us interesting perspectives regarding our vocabulary knowledge as well as the lexical resource we would need to comprehend different text genres (Nurmukhamedov and Webb, 2019).

While vocabulary assessment and vocabulary profiling could be said to be the two fields of vocabulary studies that receive the most attention, it does seem to me that research in vocabulary testing are moving way so faster that findings in vocabulary profiling are almost left behind, causing certain gap in research methodology. For example, while vocabulary tests have long employed Nation’s (2017) up-to-date BNC/COCA wordlist as the source of test items (McLean and Kramer, 2015, 2016; McLean et al., 2015; Webb et al., 2017), many recent studies on lexical coverage of texts were still stick to the Nation’s (2006) BNC wordlist (Al-Surmi, 2014; Dang and Webb, 2014; Webb and Paribakht, 2015; Nurmukhamedov, 2017; Tegge, 2017). This led to the situation where researchers who utilized these modern vocabulary tests for their studies could not reliably relate their results to the existing findings in the field. Attempts have been made to fill in the methodological gaps (Hsu, 2018; Yang and Coxhead, 2020; Nurmukhamedov and Sharakhimov, 2021), however, they are few, and certain areas of the field, including scripted and unscripted spoken discourses, remained uncovered. As a result, research on the relationship between phonological vocabulary knowledge and listening comprehension (Cheng and Matthews, 2018; Lange and Matthews, 2020; Ha, 2021b) still had to rely on the findings of Webb and Rodgers (2009a,b), which are more than 10 years old and ripe for being updated. In response to the dire need for methodologically updated findings, the present study was conducted to revisit the vocabulary demands of informal spoken English.

Table 3 presents the number of words and their proportion at each frequency level in the BNC/COCA wordlist for the SOAP, TV, and Movies corpora. Proper nouns, marginal words, transparent compounds, and acronyms were treated as separate word levels.

It is observable that around 85% of the three corpora were made up of the most frequent word families in the BNC/COCA and that nearly 90% of the corpora’s tokens were covered by the first two most frequent word families. This may be because the first two 1,000-word levels in the BNC/COCA lists primarily contain words taken from spoken corpora (Nation, 2020). The proportion of tokens showed a gradual decrease as the word frequency went down, and after the 5,000 level, the figures dropped below 1% for all the corpora, signaling the importance of high-frequency words.

Another detail that deserves attention is the proportion of proper nouns in the three corpora, which were relatively considerable, especially for the SOAP corpus. Proper nouns normally do not cause significant difficulties for reading comprehension since they can be easily recognized with the first-letter capitalization. However, concerns have been raised on the effect of proper nouns on listening comprehension (Kobeleva, 2012; Klassen, 2021). In general, auxiliary words including proper nouns (PN), marginal words (MW), transparent compounds (TC), and acronyms accounted for approximately 11% for the SOAP corpus and nearly 6% for the TV and Movies corpora.

The cumulative coverage at each word frequency level is illustrated in Table 4. At this stage, two assumptions were put forward, the first one supposes that learners did not know and could not recognize PNs, MWs, TCs, and acronyms, and the second one assumes that learners knew or could easily recognize these words.

It is obvious that without the knowledge of PN, MW, TC, and acronyms, it is impossible for viewers to achieve the minimum coverage threshold for comprehension, which is indeed worrying. However, when PN, MW, TC, and acronyms were assumed to be known, then we can once again see the optimistic scenario depicted in Webb and Rodgers (2009a,b), Tegge (2017), and Nurmukhamedov and Sharakhimov (2021). Generally, it only took 2,000–3,000 most frequent word families in the Nation’s (2017) BNC/COCA lists to cover 95% of the tokens in the corpora, and the vocabulary knowledge of the most frequent 4,000–5,000 word families was all required to reach the optimal threshold for comprehension.

Certain differences can be observed between the corpora. To be more specific, soap operas demanded the least lexical knowledge for 95% (2,000 WFs) and 98% (4,000 WFs) coverage compared to the other two. TV programs and movies share similar lexical demands when it comes to the 98% threshold. However, data from the analysis showed that the Movies corpus managed to reach the 95% coverage at the 2,000 level, while the TV programs would require a word knowledge at 3,000 level for 95% coverage. Still, it worth noting that the actual difference was really thin, approximately 0.2%, and had the tendency to become smaller as it moved down the word levels. Therefore, it is safe to state that movies and TV programs shared relatively similar vocabulary demands.

The paper revisited research on the vocabulary profiles of soap operas (Al-Surmi, 2014), TV programs, and movies (Webb and Rodgers, 2009a,b) to see if changes in sample size and research methodology would result in changes in findings. Data from the analyses showed that if the BNC/COCA wordlist were to be used as an indicator of word frequency, then the lexical demands of the researched text genres would be generally reduced. To be more specific, if learners were to base their vocabulary learning on the BNC/COCA lists, they would only need to learn 4,000 (instead of 5,000) and 5,000 (instead of 6,000 and 7,000) word families to understand 98% of the words in soap operas and movies and TV programs, respectively. It should be noted that the 1,000–2,000 word families difference could be translated into 2–4 years of English learning or even more (Webb and Chang, 2012; Ozturk, 2016).

However, the so-called “reduced lexical demands” are, in actual practice, the additional effect of the four supportive lists (PN, MW, TC, and Acronym). If we were to compare the cumulative coverage at the 3,000 and 5,000 levels, we could easily find similar figures between BNC/COCA-based and BNC-based studies. And if we were to go even further and add the proportion of the four supportive lists (PN, MW, TC, and Acronym) of this study to the cumulative coverage figures at any word level in Webb and Rodgers (2009a,b) and Al-Surmi (2014), then the same reduced lexical coverage (or even better) could be recorded. Still, this does not mean that we could simply use the BNC lists together with the four additional lists in the BNC/COCA lists. Nation (2020) introduced the BNC/COCA wordlist with clear rationales which have been proven by other researchers (Dang and Webb, 2016; Dang et al., 2020), which means that these lists were designed to work together and scholars are not encouraged to apply questionable practices to avoid unnecessary problems.

The shift in lexical profiling studies from the BNC (Nation, 2006) to the BNC/COCA (Nation, 2017) could be said to be inevitable as it harmonizes different aspects of vocabulary research. Since most vocabulary tests now utilize the BNC/COCA as the source for test items (McLean and Kramer, 2015; McLean et al., 2015; Webb et al., 2017; Ha, 2021a), it would be somehow methodologically inconsistent to relate students’ results on vocabulary test that is based on the BNC/COCA wordlist to findings of lexical studies that employed the BNC lists. Together with the development of phonological vocabulary tests (McLean et al., 2015; Ha, 2021a), teachers and vocabulary linguists can now make a reliable connection between the aural vocabulary knowledge of their students and what they can possibly understand in the real world.

The study’s findings were also in line with Nurmukhamedov and Sharakhimov (2021) which employed Nation’s (2017) BNC/COCA lists to investigate the lexical profile of English podcasts. Nurmukhamedov and Sharakhimov (2021) found that a vocabulary knowledge at the 3,000 and 5,000 levels would cover 96.75 and 98.26% of the words in English podcasts, correspondingly. These results generally suggested that learning the 5,000 most frequent word families in the BNC/COCA wordlist, an attainable learning goal for English learners in most contexts, could help learners achieve unsupported listening comprehension for informal spoken English. This claim is supported by van Zeeland and Schmitt’s (2013) study which proved that knowing 98% of the running words in a listening text would result in very high degree of listening comprehension.

The study recorded considerable proportions of PN, MW, TC, and acronyms, which aligned well with Nurmukhamedov and Sharakhimov (2021). Vocabulary linguists tend to make the assumption that these words could be easily and recognized by learners (Nation, 2006; Webb and Rodgers, 2009a,b; Nurmukhamedov and Sharakhimov, 2021). Although these assumptions could be acceptable for reading, concerns have been raised about whether it is appropriate to assume the same thing to listening (Kobeleva, 2012; Klassen, 2021). It is true that without the support of orthographic form, proper nouns or even acronyms are difficult to distinguish, and for listening-only formats like podcasts, such concerns are in evidence. However, television programs, movies, and soap operas, which are strongly supported by visualization, are different from podcasts. These non-verbal clues, such as facial expression, body gestures, and lips movements, are considered be of significant support to the processing of aural input (Harris, 2003). Such visual clues may help viewers recognize and understand PN, MW, TC, and acronyms.

This research’s findings offer updates on the lexical profiles of informal spoken English by employing up-to-date research methodology and large sample size. In general, it is evident that the BNC/COCA wordlist (Nation, 2017) would give English learners and teachers a shorter route to their intended goals. This BNC/COCA-based update also connected research findings on vocabulary profiling and vocabulary assessment, which have been in conflict for several years due to incompatible methodologies.

Despite being informative, this brief research report bears certain limitations. First, the paper revisited the findings of several studies at once, which would give readers a broad overview of the new findings and how these findings related to each other. Therefore, it was not possible for the researcher to go deeper and explore the variation in lexical coverage among texts. Future research should take a deeper look into each corpus in isolation and examine the variation in lexical demands of each text genre. Secondly, although the study showed the number of word families learners would need to achieve 95 and 98% coverage in informal spoken English. It cannot guarantee that learners would successfully comprehend a text should these requirements be satisfied. Graham (2006) showed that people may understand every single word in a text and still fail to get the general meaning, which could be due to other factors in the learners’ language proficiency and metacognitive awareness. As a result, researchers are encouraged to investigate these issues, which could possibly be done by replicating van Zeeland and Schmitt’s (2013) research methodology.

The data analyzed in this study is subject to the following licenses/restrictions: The corpora that support the findings of this study are available from Mark Davies. Restrictions apply to the availability of these corpora, which were used under academic license for this study. Data are available from https://www.english-corpora.org/ with the permission of Mark Davies. Requests to access these datasets should be directed to mark.davies@corpusdata.org.

The author confirms sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.