The data for the corpus were collected from two types of dictionaries: one paper dictionary, Oxford Advanced Learner’s English-Chinese Dictionary (OALECD); and three online dictionaries, Youdao; Oxford Learners’ Dictionary; and Collins. Their websites are:

1. http://dict.youdao.com/

2. https://www.oxfordlearnersdictionaries.com/

3. https://www.collinsdictionary.com/zh/dictionary/english

Oxford Advanced Learner’s English-Chinese Dictionary (OALECD) is created by A. S. Hornby, translated by Li Beida, and published jointly by the Commercial Press of China and Oxford University Press. The words in the dictionary are alphabetically ordered and interpreted mainly from five aspects: parts of speech, various precise English and Chinese meanings, typical examples, typical collocations, and, sometimes, necessary pictures. Its body consists of 1,774 pages. As a highly valued reference for English learners in China, it can meet their demands for communication, reading, writing and translation—it is one of the bestselling English dictionaries in China. As introduced in its publication foreword and preface, as an essential reference designed for English learners from non-English-speaking countries, its sales has been leading that of those dictionaries of the same kind. Therefore, it was selected as a key source to construct the corpus of EIVs.

Nevertheless, it should be highlighted that in this dictionary, the annotation of the transitivity of verbs is not sufficiently detailed. As the content verbs are not marked “transitive” and “intransitive,” this dictionary may cause unavoidable inconvenience in determining whether a verb should be collected or not. Therefore, three online dictionaries (Youdao, Oxford Learners’ Dictionary, and Collins) were selected to assist OALECD in determining the intransitivity of verb, as they provide explicit annotations of the intransitivity of verbs. With their assistance, the collected data were thought to be more accurate and, thus, better suited to the needs of the present study. Additionally, the meanings of verbs they provide and their illustration examples are highly detailed, which can enrich OALECD in these aspects. The three online dictionaries significantly improved the efficiency in collection of EIVs for study and in conducting an in-depth analysis of their semantic structures.

The corpus construction involved three steps: determining EIVs, creating files of EIVs, and building a Word file of EIVs. Prior to specifying the operation of each step, one point should be highlighted: What is examined was restricted to the first basic meaning of a verb and the relevant usages presented in the paper dictionary of OALECD. EIVs are used originally without direct objects; however, in extended meanings, a few of them can be used transitively. Examination of only the first basic meaning of a verb can significantly avoid the confusion of meanings and prevent the uncertainty or diversity of the lexicalization model of a given EIV.

The specific steps are as follows. The first step, critical for the corpus construction in this study, is to determine EIVs based on the newly-developed definition of pure intransitive verbs above. Specifically, the OALECD was examined alphabetically to target verbs and then the first meanings and relevant examples of the targeted verbs were investigated to speculate on their intransitivity. In case the intransitivity of a verb could not be determined, the three online dictionaries were consulted. If the verb was consistently labeled intransitive in two of them or all of them, it was collected. For example, the first meaning of the verb “doodle” on Page 430 of OALECD is “make meaningless drawings, scribbles, etc., while one is or should be thinking about something else”; it is exemplified as “Stop doodling on my notebook.” Its first basic meaning and usage suggests that it is intransitive. However, to confirm its intransitivity, the three online dictionaries were also consulted. Under the meaning that is same as, or similar to, its first basic meaning presented in OALECD, Collins demonstrates its usage by the example, “He doodled during the whole lecture”; Youdao presents “V-I (intransitive)” and the example “He looked across at Jackson, doodling on his notebook”; the Oxford Learners’ Dictionary only displays information on its usage as a noun. Considering the information provided in Collins and Youdao being similar to that in OALECD, “doodle” is deemed an intransitive verb and, thus, collected.

In developing the EIV files, a camera was used to take photos of the determined EIVs together with their meanings, examples, and attached pictures from OALECD—and where necessary, from the three online dictionaries—and stored as files named after the 26 letters on the computer.

Finally, the photos of the EIVs were copied into Word alphabetically. Meanwhile, the EIV words were input before their photos and numbered in Word so that the Word search function could be used for the EIVs needed in the study. The corpus of EIVs was hence constructed as a Word file.

The features of the constructed corpus first lie in its form. Compared to traditional well-known corpora such as the Corpus of Contemporary American English (COCA) and the British National Corpus (BNC), whose data exist in text form, the corpus established for the current study contains 710 photos and is characterized by multimodal data. Specifically, the entire Word corpus is a store of EIV photos that are alphabetically ordered in Word, rendering them easily searchable. Meanwhile, the photos present various forms of information about intransitive verbs, their first basic English and Chinese meanings and related usages, their other meanings and relevant usages, and a few attached pictures. With such multimodal data, this corpus is practical for the analysis of the semantic structures and the extraction of lexicalization models of EIVs.

Secondly, because this corpus was constructed for a particular language phenomenon (EIVs), it could be regarded as a “specialized corpus” (Zibin and Altakhaineh, 2021). In total, it comprises 602 EIVs. The specific number of EIVs is shown under each letter (Table 1).

[Motion + Patient] is the first major lexicalization model of EIVs, numbering 284 and accounting for 47.18% of the 602 EIVs. It is defined by the two major semantic components in the semantic structures of EIVs being [Motion] and [Patient]. For example, “bleed” means “lose or emit blood.” In this meaning, the part of “lose or emit” embodies the component of [Motion], and “blood,” [Patient]. Therefore, its lexicalization model is [Motion + Patient].

In addition to the two major components conflated into the semantic structures of EIVs of this type, other minor components such as [Manner], [Instrument], [Cause], [Path], [Time], [Result], [Location], and [Purpose] are also conflated in the form of their various combinations. There are 13 such combinations: [Motion + Patient], [Motion + Patient + Manner], [Motion + Patient + Instrument], [Motion + Patient + Manner + Cause], [Motion + Patient + Purpose], [Motion + Patient + Path], [Motion + Patient + Time], [Motion + Patient + Result], [Motion + Patient + Path + Manner], [Motion + Patient + Cause], [Motion + Patient + Path + Purpose], [Motion + Patient + Location], and [Motion + Patient + Path + Manner + Purpose/Time]. Their numbers and rates are shown in Table 3.

For example, the semantic structure of “proliferate” is “produce new growth or offspring rapidly.” In the structure, “produce” embodies the semantic component of [Motion]; “new growth or offspring” specifies the component of [Patient]; “rapidly” exemplifies the component [Manner]. Therefore, the lexicalization model of “proliferate” is [Motion + Patient + Manner]. Similarly, “cough” means “send out air from the lungs violently and noisily, esp. to clear one’s throat or when one has a cold, etc.” In this structure, the parts of “send out,” “air,” “from the lungs,” “violently and noisily,” and “esp. to clear one’s throat or when one has a cold” embody, respectively, [Motion], [Patient], [Path] (fusing the starting point), [Manner], and [Purpose/Time]. Therefore, its lexicalization model is [Motion + Patient + Path + Manner + Purpose/Time].

In this model the specific parts embraced in the semantic structures of EIVs mainly embody the semantic components of [Motion] and [Manner]. For instance, in the meaning of “flounder,” “move or struggle” embodies [Motion] and “helplessly or clumsily” embodies [Manner]. As Table 3 shows, this model numbers 168; it is the second most common, immediately after the [Motion + Patient] model.

There are six subtypes: [Motion + Manner], [Motion + Manner + Cause], [Motion + Manner + Purpose], [Motion + Manner + Location], [Motion + Manner + Path], and [Motion + Manner + Path + Instrument]. This demonstrates that in addition to the two major components of [Motion] and [Manner], the semantic structures of EIVs of this type incorporate the components of [Cause], [Purpose], [Location], [Path], and [Instrument]. For example, “jump” is structured by [Motion] of “move,” [Manner] of “quickly,” [Path] of “off the ground, etc., esp. up into the air,” and [Instrument] of “by using the force of the legs and feet”; “equivocate” is semantically constructed by [Motion] of “speak,” [Manner] of “in an ambiguous way,” and [Purpose] of “to hide the truth or mislead people.”

In this model, the two significant components embodied by the parts in the semantic structures of EIVs are [Motion] and [Path], with other components of [Manner], [Cause], [Patient], and [Result] occurring in certain combinations. In total, it numbers 54 and accounts for 8.97% of the 602 EIVs.

Table 2 shows its seven sub-models: [Motion + Path], [Motion + Path + Manner], [Motion + Path + Cause], [Motion + Path + Manner + Cause], [Motion + Path + Patient + Path], [Motion + Path + Patient + Manner], and [Motion + Path + Result]. For example, “emerge” is semantically constructed by [Motion] of “come” and [Path] of “out or up (from water, etc.),” whereas “flinch,” is semantically constructed by [Motion] of “move or draw,” [Path] of “back,” [Manner] of “suddenly,” and [Cause] of “from shock, fear or pain.”

This model implies that the meaning of the EIV is constituted mainly by the parts embodying the components of [Motion] and [Result]. For example, in “deteriorate,” the part “become” embodies [Motion] and “worse in quality or condition” embodies [Result]. This model numbers 49 and occupies 8.14% of the 602 EIVs.

Additionally, in the semantic structures of certain EIVs of this type, other components of [Cause], [Manner], and [Time] are added to [Motion + Result]. Therefore, the other sub-models of [Motion + Result + Cause], [Motion + Result + Manner], and [Motion + Result + Time] occur. Let us consider “shrink,” in which [Cause] of “because of moisture or heat or cold” is conflated into its semantic structure in addition to [Motion] of “become” and [Result] of “smaller.” Hence, its lexicalization model is [Motion + Result + Cause].

In this model, which numbers 21 and occupies 3.49% of the 602 EIVs, [Motion] and [Location] are the two major semantic components embodied in the semantic structures of EIVs. For instance, the semantic structure of “encamp” covers [Motion] of “settle” and [Location] of “in camp.”

Like the lexicalization models above, this model also embraces sub-models: [Motion + Location], [Motion + Location + Instrument], [Motion + Location + Time], and [Motion + Location + Cause]. For example, “levitate” is semantically constructed by [Motion] of “rise and float,” [Location] of “in the air,” and [Instrument] of “esp. by means of supernatural powers.”

This model contains [Motion + Purpose] and [Motion + Purpose + Time]. It only numbers six and accounts for about 1% of the 602 EIVs. For example, the semantic structure of “contend” includes [Motion] of “struggle” and [Purpose] of “in order to overcome a rival, competitor, or difficulty,” and that of “rally” conflates [Motion] of “come together,” [Purpose] of “to make new efforts,” and [Time] of “after a defeat or when there is danger, need, etc.”

The other types of lexicalization models of EIVs are [Motion], [Motion + Time], [Experiencer + Motion], [Motion + Cause], and [Patient + Motion + Agent]; they number, respectively, five, five, five, four, and one, and account for, respectively, 0.83%, 0.83%, 0.83%, 0.66%, and 0.17% of the 602 EIVs.

Among these, the model of [Motion] implies that the meanings of EIVs only conflate the part embodying [Motion]. For example, “fare” is semantically constituted only by [Motion] of “progress.” The [Motion + Time] model refers to the case in which the semantic constructions of EIVs conflate the parts embodying [Motion] and [Time]. For example, the meaning of “coincide” fuses [Motion] of “occur” and [Time] of “at the same time.” The [Experiencer + Motion] model may be typically exemplified by “blow.” In its semantic structure, the part of “the air” embodies [Experiencer] and “moves” embodies [Motion]. Regarding the remaining two models, typical examples are “writhe” and “depend.” In the semantic structure of “writhe,” the part of “twist or roll about” embodies [Motion] and “because of great pain” embodies [Cause]. In “depend,” “the first thing” is [Patient], with “will be affected or determined” and “by the second” being [Motion] and [Agent], respectively.

As Tables 2, 3 show, nearly half of the EIVs are lexicalized with the [Motion + Patient] model. In other words, [Patient] is the semantic component second only to [Motion] being conflated into the semantic structures of EIVs. According to Li (2006, p. 197), [Patient] refers to “the entity undergoing the effect of some action, often undergoing some change in state.” Although [Patient] does not strictly correspond to the syntactic object, Dowty’s (1991) Argument Selection Principle states that “the argument having the greatest number of Proto-Patient entailments will be lexicalized as the direct object.” In other words, the typical patient usually occupies the syntactic position of object.

Accordingly, it might be inferred that as [Patient], which usually occurs at the syntactic position of object, has been incorporated into the semantic structures of EIVs, the given EIVs need not take objects syntactically, as doing so would lead to semantic redundancies in their syntactic structures. This is because the incorporated [Patient], functioning as invisible objects, makes the meanings of EIVs relatively independent and complete. Take “bleed” and “sniff” as examples. As Table 3 shows, their semantic structures are “lose or emit blood” and “draw air,” respectively. With the conflation of [Patient] of “blood” and “air,” their meanings have been relatively independent and complete. If “blood” and “air” are placed again after “bleed” and “sniff” in the clauses, “His head had struck the sink and was bleeding” and “She wiped her face and sniffed loudly,” the two clauses would be semantically redundant and thus, usually, unacceptable.

There might be some confusion here. Those EIVs that do not incorporate [Patient] also do not take objects initially. It appears that there is no [Patient] immediately after the component of [Motion] in their lexicalization models such as [Motion + Manner], [Motion + Path], [Motion + Purpose], and [Motion + Location], but a further analysis of verbs embodying [Motion] in these models will provide some clues. Typical EIVs such as “equivocate,” “commute,” “emerge,” “contend,” and “levitate” have been selected to illustrate the issue.

As Table 3 shows, the major lexicalization models of “equivocate” and “commute” are [Motion + Manner]. In their meanings, the parts representing [Motion] are two verbs: “speak” and “travel.” The specialized corpus of EIVs of the present study shows that “speak” initially means “utter words” and “travel,” “make a journey,” thereby falling under the [Motion + Patient] model. This demonstrates that they themselves are intransitive verbs. Based on this analysis, it may be inferred that in the semantic structure of a given verb, if the part embodying the constant semantic component of [Motion] is an intransitive verb, the given verb will probably also be intransitive because it inherits the intransitive feature of the intransitive verb used to explain it (see Figure 3). This given verb is EIV2 and the verb used to explain it is EIV1 in Figure 3.

This figure shows that EIV2 does not syntactically take a direct object because its meaning has indirectly incorporated the part specifying the constant semantic component of [Patient] originating from the semantic structure of EIV1. Therefore, the two EIVs, “equivocate” and “commute” are used without objects in the sentences, “Do not equivocate with me—I want a straight answer to a straight question” and “She commutes from Oxford to London every day,” just as shown in the specialized corpus of EIVs.

As for “emerge,” “contend,” and “levitate,” their lexicalization models are [Motion + Path], [Motion + Location + Instrument], and [Motion + Purpose], respectively. As Table 3 shows, in their semantic structures, [Motion] is embodied, respectively, by “come,” “struggle,” and “rise and float,” which themselves are intransitive according to the specialized corpus of EIVs. When employed to explain “emerge,” “contend,” and “levitate,” their intransitive feature might be inherited by the latter. This inheritance of intransitivity may be the reason the latter do not take objects in usages such as “The swimmer emerged from the lake,” “Several teams are contending for the prize,” and “He claims that he can levitate.”

The analysis of the [Motion + Result] model shows that the actions expressed by EIVs of this model are internalized rather than externalized. They only involve the inner changes of states of entities and will not exert any influence on other entities, thereby syntactically taking objects out of necessity. This is consistent with Rappaport Hovav and Levin’s (1998) account of the verbs of internally caused change of state (such as “grow” and “wilt”) as different from those of externally caused change of state (such as “open” and “break”). The typical verb used to express the semantic component of [Motion] of this model is “become.” Let us consider as examples the EIVs “deteriorate” and “shrink.” Their semantic structures of “become worse in quality or condition” and “become smaller” show the inner changes of state of entities. Accordingly, they are used without objects, as in, “Leather can deteriorate in damp conditions” and “The dough shrank in the cold air.”