Constructions in competition: The development of the impersonal verb hunger and the adjectival periphrasis be hungry in Early Modern English

ABSTRACT The present study is concerned with the syntactic and semantic development of the impersonal verb hunger in Early Modern English. An analysis of corpus data has been carried out on ca. 20 million words drawn from EEBOCorp 1.0 (1473–1700). Results show that, from a semantic perspective, the verb hunger undergoes a process of metaphorical extension involving a change from the original meaning ‘to feel hunger’, in the domain of Physical Sensation, to the meaning ‘to desire’, in the domain of Emotion. In this latter sense, the verb becomes predominantly associated with prepositional complements (e.g. 1542, our hungry soules [...] hunger for y^ word of God). Also in the course of the Early Modern period, the verb is subject to competition with the adjectival periphrasis be hungry, especially in the sense ‘to feel hunger’. The article concludes by putting forward hypotheses to explain the motivations for these various developments.


Introduction
This article is concerned with the historical development of the formerly impersonal verb hunger (OED s.v. hunger, v). In earlier English, hunger could occur in a special type of morphosyntactic pattern known as 'impersonal construction', which is characterised by the lack of a grammatical subject controlling verbal agreement, as illustrated in example (1) below. 1 (1) Se ðe cymes to me ne hyncgreð hine.
'Blessed are they that hunger and thirst for [i.e. desire] righteousness' (1382 J. Wyclif Matt. v. 6;OED s.v. hunger, v. † 4.) The competition between the verb hunger and the adjectival construction be hungry remains unexplored in the literature. This is largely a consequence of the fact that the development of impersonal verbs during the Early Modern English (henceforth EModE) period has received very little attention overall, with the focus being placed on OE and ME (see Elmer 1981;Allen 1995;Möhlig-Falke 2012;Miura 2015). In this regard, Miura (2015: 9) points out that 'a large corpus-based study of not just Old and Middle English but also early Modern English is indispensable' (see also Möhlig-Falke 2012: 236). The historical interest of EModE is no doubt very great, since this period comprises the span of time when impersonal verbs were in the process of adjusting their syntax to the new possibilities of the language, after the morphosyntactic transformations that took place during OE and ME had been completed.
In this context, the present paper aims to analyse the syntactic and semantic development of the verb hunger in EModE paying special attention to its decline in frequency and the parallel surge of the adjectival construction be hungry. For this purpose, a body of data has been extracted from Early English Books Online Corpus 1.0 (henceforth EEBOCorp 1.0), a 525-million-word corpus covering the period 1473-1700. This has been complemented by a comprehensive survey of the Oxford English Dictionary (OED) and Middle English Dictionary (MED) entry quotations. More specifically, the goals are: 1) to carry out a diachronic analysis of the senses conveyed by hunger and be hungry; 2) to describe the frequency of occurrence of the complementation patterns found with the verb and with the adjectival construction; 3) to draw a comparison between hunger and be hungry in terms of their syntactic and semantic behaviour. On the basis of the data obtained from the above-mentioned sources, this paper further seeks to answer two main research questions concerning: 1) the motivations behind the semantic extension of the verb hunger from the physical-sensation sense 'to feel hunger' to the emotion sense 'to desire'; and 2) the motivations behind the competition between hunger and the adjectival construction be hungry for the expression of the physical-sensation sense 'to feel hunger'.
Results are interpreted within the theoretical framework of Cognitive Grammar (Fillmore 1977;1986;Langacker 1987;Croft 1991; and Construction Grammar (Goldberg 1995;2006). Special reference will be made to the notion of 'forcedynamics' -i.e. 'how entities interact with respect to force' (Talmy 2000: 409) -which is of value to describe, not only the physical transmission of force between a verb's participants (Croft 1991: 162), but also the mental relationship between the participants of emotional processes (Croft 1991: 166-167;2012: 199).
The remainder of the paper is organised as follows. Section 2 is concerned with the various paths of development followed by impersonal verbs after impersonal constructions were lost. Section 3 describes the data and methodology used. Section 4 considers the origin and syntactic patterning of the verb hunger based on a survey of the OED and MED entry quotations. Section 5 presents the results of the analysis of corpus data focusing, firstly, on the EModE verb hunger, and, secondly, on the adjectival construction be hungry. This is followed by a discussion of findings in Section 6, and by a summary of the main findings and conclusions in Section 7.

The development of impersonal verbs in the history of English
Previous studies have shown that impersonal verbs vary extensively as to the course of development they followed after the loss of impersonal constructions (e.g. Fischer & van der Leek 1983;Möhlig-Falke 2012). In some cases, impersonal verbs disappeared from the language altogether (e.g. ME bōten 'avail' or reusen 'repent', Möhlig-Falke 2012: 15). In other cases, impersonal verbs were replaced by non-impersonal verbs having no etymological relation (e.g. ME, me reweþ '[there] is regret [to] me' gave way to PDE, I regret, Castro-Chao 2021b). In yet other cases, impersonal verbs survived as personal verbs beyond the EModE period, as is the case of ache, ail, hunger, long for, rue or thirst, among others (Möhlig-Falke 2012: 205).
The paragraphs that follow summarise the syntactic paths available to impersonal verbs after the general loss of impersonal constructions. The account provided is adapted from the works of Fischer & van der Leek (1983: 365-366) and Möhlig-Falke (2012: 217-218).

Path I:
The Experiencer argument is interpreted as subject, whereas the second argument of Stimulus, if expressed at all, is encoded as object. This path corresponds to so-called Experiencer-subject constructions (Fischer & van der Leek 1983: 352-354) and it is the path followed by the verb hunger, as well as by predicates such as like, need or thirst; e.g. PDE, She [=Experiencer, subject] likes money [=Stimulus, object] (example adapted from Fischer & van der Leek 1983: 363).

Path II:
The Experiencer is interpreted as object and the Stimulus is encoded as the subject of the clause. This path corresponds to so-called Experiencer-object constructions (Croft 1991: 219;cf. Fischer & van der Leek 1983: 349-352). It is the path followed by impersonal verbs such as ail or please; e.g. PDE, Her decision [=Stimulus, subject] pleased me [=Experiencer, object] (example adapted from Fischer & van der Leek 1983: 363).

Path III:
The verb is used in a (h)it-extraposition construction, with the place of the subject filled by formal (h)it, and the Experiencer realised as a (prepositional) object. The Stimulus is often realised by an extraposed clausal complement. This path was followed by please or seem; e.g. PDE, It seemed (to him) that the weather would not last (example adapted from Fischer & van der Leek 1983: 363).

Path IV:
The impersonal verb is used in a middle-reflexive pattern in which the Experiencer is pleonastic, being encoded both as the subject and as a co-referential reflexive pronoun; e.g. OE, He sceamode him 'He was ashamed of himself' (example adapted from Möhlig-Falke 2012: 189, my translation).

Path V:
The impersonal verb is used in passive or adjectival patterns formed by the combination of to be or to become plus a past participle or a related adjective, with the Experiencer argument being encoded as the subject of the clause. The Stimulus may be optionally encoded as an adjunct. Verbs that appear in passive or adjectival patterns include hunger, please or thirst; e.g. 1992, Daddy wasn't any too pleased about it [. . .] (OED s.v. please, v. 4. a.;see also Croft 1991: 216;Wierzbicka 1999: 302;Möhlig-Falke 2012: 223-225).
As hinted at in Section 1, the development of hunger follows Paths I and V above. That is, the verb adopted personal syntactic uses in which the Experiencer is expressed as subject (e.g. example (2), Section 1: I hungerd), though in PDE the adjectival construction has come to be largely preferred (see example (3), I was hungre) in some functions. As pointed out by Möhlig-Falke (2012: 224), the verb hunger has 'been retained in English in marginal use as [a] stylistically marked' alternative to the more neutral adjectival construction. In other words, the verb survives in PDE only as an archaic usage, mostly associated with religious contexts such as PDE when he [i.e. Jesus Christ] had fasted forty days and forty nights, he afterwards hungered (Lexico's Dictionary s.v. hunger verb 2); by contrast, adjectival constructions are found in more stylistically-neutral contexts -e.g. PDE, I was feeling ravenously hungry (Lexico's Dictionary s.v. hungry adjective 1).
From a semantic perspective, adjectival constructions -i.e. Path V, example (3), I was hungre -show the particularity that they assign a stative and uncontrolled interpretation to the State of Affairs (henceforth SoA; see Wierzbicka 1999: 302;further Möhlig-Falke 2012: 193). By contrast, the verb hunger, when used in the active voice -i.e. Path I, example (4), Blessid be þei þat hungren [. . .] riȝt~wisnesse -assigns a dynamic and to some extent volitionally controlled interpretation to the SoA (see Wierzbicka 1999: 302;Castro-Chao 2021b). The term 'State of Affairs', as employed here, is adopted from Möhlig-Falke (2012: 56;also Dik 1997: 51;Van Valin & LaPolla 1997: 83), and refers to the verb's ability to denote: something that can be said to occur, take place, or obtain in some world; it can be located in time and space; it can be said to take a certain time (have a certain duration); and it can be seen, heard, or otherwise perceived. (Dik 1997: 51)

Data and methodology
The present investigation is based on a large bank of corpus data. The need for a corpusbased investigation on the development of impersonal verbs in EModE is made evident by the most recent studies on impersonals (see in this regard Miura 2015: 96; also Möhlig-Falke 2012: 30).
This article draws on the methodology adopted in previous work by Castro-Chao (2019: 136-137;2021a, 2021b. A random sample of 891 texts has been extracted from the 525 million words contained in EEBOCorp 1.0 (1473-1700), 2 in order to obtain a manageable subcorpus which totals ca. 20 million words. This subcorpus is structured into four sections of 50-year subperiods of comparable size, covering the entire EModE period (1500-1700). The distribution of the number of texts and the wordcount summary per 50-year subperiod are displayed in Table 1.
A dataset of 331 examples has been retrieved from these 891 texts: 193 tokens for the verb hunger and 138 for the adjectival construction be hungry. Firstly, possible spellings for both the verb and the adjective were identified from the OED corresponding entries (i.e. s.v. hunger, v. Forms and hungry, adj. Forms). Subsequently, a word list was generated using the concordancer AntConc (Anthony 2019), so as to identify the array of spellings and forms attested in the subcorpus extracted from EEBOCorp 1.0. These were then used to retrieve the examples with AntConc. For the purposes of this paper, the dataset has been categorised according to the following four parameters: 1) 'subperiod'; 2) 'verb sense'; 3) 'complementation pattern'; and 4) 'type of preposition'.

Origin and syntactic patterning of hunger
In this section, I look at the origin and development of hunger, from OE hyngrian (> ME hungren, see Section 1), based on the OED and MED entry quotations. The discussion considers first the semantic development of hunger, followed by a brief discussion of the event structure of each verb sense. Subsequently, an account is provided of the range of complementation patterns documented with this verb in the OED and MED. The verb hunger may denote three different senses: (i) 'to feel hunger', illustrated in example (2) above, Section 1; (ii) 'to desire with longing' (henceforth 'to desire'), illustrated in (4), Section 1; and (iii) 'to subject to hunger', illustrated in (5) below. The senses that will be taken into account for present purposes are sense (i) 'to feel hunger' and sense (ii) 'to desire', as these are the ones attested in the EModE data, as shall be explained in Section 5.1 below. The sense 'to feel hunger' is documented since OE in both impersonal and personal use (see OED s.v. hunger, v. †1. and 2., respectively), and it represents the earliest attested use of the verb. In PDE usage this verb sense has become 'archaic', as is explicitly pointed out by Lexico's Dictionary (s.v. hunger verb 2), and as is implicit from the OED, whose last quotation in this sense dates from 1881 in a passage from the Bible (R.V.) Matt. iv. 2, the previous quotation going back to 1794. The sense 'to desire' constitutes a metaphorical extension of the literal sense 'to feel hunger', from hunger for PHYSICAL NOURISHMENT to hunger for SPIRITUAL NOURISHMENT (see, e.g. Mosedale 1997) 3 ; it emerges later than this and is first attested in the MED in the late 14 th century (s.v. hungren v. 1. (c)). An earlier example from late OE is available in the OED (s.v. hunger, v. †4.), but it is documented in a text that is dependent on Latin (West Saxon Gospels), and its reliability for reflecting a genuine use of this verb sense is therefore questionable. The sense 'to subject to hunger', for its part, is attested since the 16 th century. Given that, as already pointed out, the verb hunger is subject to competition with the adjectival construction be hungry (see Sections 1 and 2), mention should be made of the latter as documented in the Dictionary of Old English (DOE), OED and MED. It is first attested in OE in the literal sense 'to feel hunger' (DOE s.v. hungrig adj. 1; see also OED s.v. hungry, adj. 1. a.; MED s.v. hungrī(e adj. 1. (a)). In the metaphorical sense 'to desire', the adjectival pattern is not documented in the OED (s.v. hungry, adj. 4. a.) and is recorded just once in the extensive MED quotation database (s.v. hungrī(e adj. 1. (a), a1438, my sowle is euyr a-lych hungry 'my soul is always terribly hungry'); but the attributive use of the adjective hungry in the sense 'having a strong desire or craving (for, †after, †of anything)', illustrated in (6), is found since the 13 th century.
(6) For he [. . .] Allen 1995: 71). This means that, in a sentence like 1340 I hungerd and yhe me fedde 'I was hungry and you fed me' (example (2)), it is irrelevant to clarify the identity of what is needed. Rather, it is assumed by the speaker and the addressee that the Needed corresponds to some indefinite edible substance which can satisfy the physical feeling of hunger (see Fillmore 1986: 96;Goldberg 1995: 58).
In the sense 'to desire', hunger belongs to the semantic domain of Emotion rather than Physical Sensation (Möhlig-Falke 2012: 90-93). It profiles two distinct semantic roles in the semantic frame <Desirer Desired>, where both semantic roles are lexically profiled. Thus the process of metaphorical extension involves, on the one hand, a transference from the semantic domain of Physical Sensation to the domain of Emotion, and, on the other, a change in the verb's valency from one (in the literal sense 'to feel hunger') to two arguments (in the metaphorical sense 'to desire').
From the perspective of force-dynamics (Talmy 2000), the literal and the metaphorical senses show notable differences as well. Thus, the event denoted by the physicalsensation sense 'to feel hunger' lacks dynamicity in that it is conceived of as a state rather than a process. This is because a dynamic relationship holding between two distinct participants -namely an Initiator and an Endpoint -is missing (Croft 1991: 173;Möhlig-Falke 2012: 57;Castro-Chao 2021b). For instance, notice that in an example like PDE when he [i.e. Jesus Christ] had fasted forty days and forty nights, he afterwards hungered (Section 2), he does not represent an Initiator which starts or stops the sensation of hunger in a controlled manner. Hence, the SoA denoted by the verb may be said to be uncontrolled, without being directed from an intentional Initiator towards an Endpoint which is in turn affected by it (see Croft 1991: 166-167;2012: 197-208). Instead, he represents an Endpoint rather than an Initiator, since its referent turns out to be the one that is affected by the internal change of state caused by the feeling of hunger (cf. Möhlig-Falke 2012: 58).
In contrast, the metaphorical sense 'to desire' denotes a process rather than a state, and it conversely profiles a dynamic relationship between an Initiator and an Endpoint. Thus, in an example like he hungered for a sense of self-worth (Lexico's Dictionary s.v. hunger verb 1), the Desirer, he, represents the Initiator in that he is viewed as responsible for directing his attention towards the object of desire in a controlled manner -i.e. towards a sense of self-worth (cf. Croft's category of 'mental activity verbs ', 1991: 217). The Desired, in turn, represents an Endpoint insofar as it is the entity to which the SoA is directed (see Castro-Chao 2021b; further Allen 1995: 144; Möhlig-Falke 2012: 92). Note, however, that the Desired turns out to be totally unaffected by the SoA, since the Desirer's inner state does not extend or have any visible effect on the second participant, i.e. a sense of selfworth.
With regard to the complementation patterns of hunger, this verb is historically documented in impersonal use (e.g. example (1)) and personal use (e.g. example (2)). Impersonal constructions are attested from OE through to the late 15 th century (see OED s.v. hunger, v. †1.). Impersonal and personal usages coexist over the same period and, once the impersonal usage is lost, the personal one is left as the only one surviving into PDE (see OED s.v. hunger, v. 2. and 3.). In personal use, hunger can govern three different types of complements, listed below by order of first attestation.

The verb hunger in EModE
This section presents the syntactic and semantic analysis of the verb hunger in EModE. The discussion, as will be seen below, also pays attention to the collocates which are most commonly found with each of the verb senses. Table 2 shows the diachronic evolution of senses distributed by 50-year subperiod, and by order of overall frequency. Figure 1 provides the relative frequencies, also distributed across 50-year subperiods. The label 'overlapping' is adopted from Robinson's work on semantic change and lexical polysemy (2010; see also Núñez-Pertejo 2017), and is reminiscent of the notion of 'bridging context' employed in grammaticalisation research (see Diewald 2002;Heine 2002). 'Overlapping' is used here to accommodate examples where one cannot clearly distinguish between the sense 'to feel hunger' and the sense 'to desire' (see example (12) further down), as a result of the nondiscreteness of semantic categories. In turn, the label 'other' has been used for one instance which could not be classified due to the omission of crucial information in the source text.
The data presented in Table 2 and Figure 1 suggest that in the EModE period the verb is preferably used to convey the sense 'to desire', with over half of the total number of  occurrences (52.3%, 101 tokens). This is followed by the sense 'to feel hunger', with 42.5% (82 tokens), while the frequency of the overlapping sense is very low at 4.7% (9 tokens). From a diachronic perspective, an increase can be observed in the sense 'to desire' from 36.6% in S1 (26 tokens) to 41.7% in S2 (15 tokens) and 75% in S3 (42 tokens). In S4 the value drops to 60% (18 tokens), although a tendency to increase is still at work, since the value of S4 is lower than the value of S3 -i.e. 75% -but higher than that of S2i.e. 41.7%. In parallel, the literal sense 'to feel hunger' decreases from 56.3% in S1 (40 tokens) to 47.2% in S2 (17 tokens) and 23.2% in S3 (13 tokens). In S4 the value rises back up to 40% (12 tokens) -i.e. 16.8% higher than the value of S3 -but is still 7.2% lower than that of S2, and lower than the value of the sense 'to desire'. As for overlapping examples, all of these belong to S1 (5.6%, 4 tokens) and S2 (11.1%, 4 tokens), naturally the periods when changes are taking place, but in S3 there is just 1 token (1.8%), and in S4 there is none. Example (10) below illustrates the sense 'to desire', with the prepositional complement referring to the abstract concept y^ word of God. Example (11) illustrates the sense 'to feel hunger', with the unprofiled Needed unexpressed. As for the overlapping example illustrated in (12), the prepositional complement makes reference to breade, and thus involves a substance which is susceptible of satisfying the feeling of physical hunger. It is noteworthy that the sense 'to subject to hunger' is unattested in the EModE data examined, despite the fact that it is documented in the OED from the 16 th century onwards (see Section 4). Overlapping examples such as (12) deserve special mention on the grounds that the verb shows properties of both of its two main senses, i.e. 'to feel hunger' and 'to desire'. What happens in such overlapping examples is that the Feeler (or Desirer) is said to feel a physical desire for an edible substance -i.e. breade in example (12) -which in the context of Christian doctrine is metaphorically interpreted as 'spiritual salvation'. In line with this, overlapping uses are found in collocation with nouns such as bread (3 tokens, example (12)), food (1 token) or meat (1 token), all of which refer to concrete edible subtances which correlate with physical hunger, but which represent a spiritual state whereby the soul longs for union with God. The metaphorical sense 'to desire' stands in contrast with overlapping examples in that it collocates with nouns which denote abstract concepts, and are totally unrelated to the original domain of Physical Sensation: e.g. justice (4 tokens) or righteousness (28 tokens), and often refer to divine entities like Christ (7 tokens), God (1 token), heaven (2 tokens) or (the) word (of God) (7 tokens, example (10)).
Given that overlapping examples involve contexts in which a double interpretation is possible, they resemble Heine's notion of 'bridging contexts'. This notion is employed in grammaticalisation research for contexts which 'trigger an inferential mechanism to the effect that, rather than the source meaning [i.e., our literal sense "to feel hunger"], there is another meaning, the target meaning [i.e., our overlapping sense "to feel hunger for/ desire spiritual salvation"], that offers a more plausible interpretation of the utterance concerned; however, an interpretation in terms of the source meaning cannot be ruled out' (Heine 2002: 84-86). In turn, the purely metaphorical sense 'to desire' illustrated in (10) is analogous to so-called 'switch contexts', in which '[t]here is a new context which is incompatible with the source meaning' (Heine 2002: 86).
Also relevant here are the kinds of discourse contexts in which the verb hunger is found. These very frequently are religious contexts such as those illustrated in (10)-(12) above, an association which becomes apparent also in the OED entry quotations, where 19 out of 29 quotations are documented in religious texts. 6 See in this regard examples (1), (2), (7) and (9), all of which are drawn from well-known religious works like the Lindisfarne Gospels, Prick of Conscience, the West Saxon Gospels and Tyndale's translation of the Bible. We may thus infer that it is in such types of texts that the sense 'to feel hunger' develops the figurative interpretation 'feel hunger for/desire spiritual salvation' illustrated in the overlapping example (12).
In other words, it is in the context of Christian doctrine that a longing for spiritual salvation is commonly expressed in terms of a physical lack of food, an interpretation which becomes explicit in example (13) below in which this bread of life clearly alludes not only to the physical nature of bread -(as well as wine) as present in the Christian Eucharist -but to its symbolic nature as the spiritual salvation that comes from taking the body of Christ.
(13) How be it, too the intent we may hunger after this bread of life, and thirste after this heauenly drinke, as this multitude did. (Niels Hemmingsen, 1569, A Postill, or, Exposition . . .) Moving now to the syntactic context, Table 3 below shows the raw frequencies for each of the complementation patterns of hunger documented in the corpus, arranged in terms of each of the verb senses with which they correlate -i.e. literal, metaphorical and overlapping. Percentages are given in brackets. Figure 2 below displays the relative frequencies of each complementation pattern, distributed across 50-year subperiods. The results presented above reveal that the verb hunger shows an overall preference for prepositional complements (52.8%, 102 tokens), as in example (14). Second in frequency comes one-place complementation, with 41.5% of uses (80 tokens), for which see example (15). NP complements are much less frequent, with 3.6% of occurrences (7 tokens, e.g. example (16)), whereas clausal complements occur in only 1.6% of cases (3 tokens, e.g. example (17)). Notice that clausal complements are not attested in the OED entry quotations (see Section 4).  From a functional perspective, the analysis of corpus data reveals that one-place complementation regularly correlates with the sense 'to feel hunger', as in examples (11) and (15). In terms of Fillmore's typology of null complements (1986: 96), this sense may be considered as having an indefinite null complement: the second participant is left unexpressed because to specify its identity is irrelevant in the context. As explained in Section 4, the Needed in the sense 'to feel hunger' is 'conceptually nonsalient [in that it is] unknown or unspecific' (Möhlig-Falke 2012: 89; cf. Allen 1995: 71); it is therefore natural that this argument is syntactically unexpressed, since it is not semantically required to complete the verb's lexical meaning (see Levin 1993: 33;Goldberg 1995: 56-59; Castro-Chao 2021b). By contrast, the patterns which have an explicit complement -i.e. prepositional, NP and clausal patterns -correlate either with the metaphorical sense 'to desire' or with the overlapping sense 'to feel hunger for/desire spiritual salvation', but not with the original univalent sense 'to feel hunger'. In the case of the overlapping sense, the Needed becomes syntactically explicit because reference needs to be made to the edible substance which is figuratively interpreted as 'spiritual salvation' -see example (12). As to the sense 'to desire', the Desired is made explicit simply because this participant role forms part of the semantic profile of the verb <Desirer Desired> (Section 4), and its expression is therefore required to complete the verb's lexical meaning.

(14) O when a man hungers for
From a diachronic perspective, it stands out that prepositional complements experience an increase from 33.8% in S1 (24 tokens) to 50% in S2 (18 tokens) and 76.8% in S3 (43 tokens). In S4 the frequency unexpectedly declines down to 56.7% (17 tokens), but notice, however, that an increasing progression can still be discerned, since the value of S4 -i.e. 56.7% -is higher than that of S2 -i.e. 50%. As for one-place complementation, this undergoes a parallel decrease from 54.9% in S1 (39 tokens) to 47.2% in S2 (17 tokens) and 21.4% in S3 (12 tokens); yet again, the frequency turns to the opposite direction in S4, rising to 40% (12 tokens) -i.e. 18.6% higher than S3. NP complements are attested in all four subperiods, but they remain below 6% of uses throughout EModE. Lastly, clausal complements are attested only in S1 (4.2%, 3 tokens).

The adjectival construction be hungry in EModE
This section presents the syntactic and semantic analysis of the adjectival construction be hungry. As for the verb, the discussion pays attention as well to the noun collocates found in each of the senses conveyed by the construction. Table 4 shows the diachronic evolution of senses throughout EModE, arranged by order of frequency; Figure 3 displays the relative frequencies distributed across 50-year subperiods.
Overall, the predominant sense of be hungry is 'to feel hunger', with 92.8% of uses (128 tokens); this is followed by a very low 4.3% of uses in the sense 'to desire' (6 tokens) and 2.9% in the overlapping sense 'to hunger for/desire spiritual salvation' (4 tokens). As a first observation, we can see that this is in clear contrast with the data for the verb hunger, which conversely predominates in the sense 'to desire' (see Table 2 and Figure 1). From a diachronic perspective, the sense 'to feel hunger' remains predominant with the adjectival periphrasis throughout the entire EModE period, with all four subperiods standing above 85% of uses. In contrast, the sense 'to desire' remains highly infrequent overall, standing at or below 10% of uses in all four subperiods.
Example (18) below illustrates the sense 'to feel hunger', with the unprofiled Needed unexpressed, and example (19) illustrates the sense 'to desire', with the prepositional complement referring to the abstract concept knowledge (1 token), which is unrelated to the domain of Physical Sensation. Other collocate nouns found in the sense 'to desire' are gain (2 tokens) or righteousness (3 tokens). As to the overlapping sense 'to hunger for/desire spiritual salvation', illustrated in (20), the prepositional complement refers to fode (2 tokens), and hence involves a substance  which is susceptible of satisfying the feeling of physical hunger (cf. example (12)  The data further suggest that the construction be hungry competes with the verb hunger in the domain of Physical Sensation (sense 'to feel hunger'), but not in the domain of Emotion (sense 'to desire'). The diachronic competition between hunger and be hungry in the domain of Physical Sensation is reflected in Table 5 and Figure 4; Table 6 and Figure 5 show the distribution in the domain of Emotion. The data in Table 5 and Figure 4 suggest that the expression of the literal sense 'to feel hunger' is preferred with the adjectival construction, which rises from 47.3% in S1 (35 tokens) to 66.7% in S2 (34 tokens) and 77.4% in S3 (41 tokens); as previously noted, the trends in S4 tend to reverse and here, the frequency recedes to 60% (18 tokens), yet higher than in S1. By contrast, the verb tends to lose frequency in the sense 'to feel hunger', decreasing from 52.7% in S1 (39 tokens) to 33.3% in S2 (17 tokens) and 22.6% in S3 (12 tokens); in S4 the frequency rises back up to 40% (12 tokens), but remains lower than in S1. In contrast, Table 6 and Figure 5 show that in the sense 'to desire' it is the verb that is clearly dominant, as it stands at or above 90% of uses across all four 50-year subperiods. This means that the verb declines in parallel to the rise of the adjectival construction be hungry only in the literal sense 'to feel hunger', but not in the metaphorical sense 'to desire'.
The fact that the verb remains predominant in the sense 'to desire' is in accordance with the data presented in Table 2 and Figure 1, which reveal that hunger becomes increasingly associated with the expression of 'to desire', while losing ground in the expression of 'to feel hunger'. Similar findings have been reported in Castro-Chao (2021b) with reference to the adjectival construction be thirsty, which follows a parallel path of development in that 'adjectival patterns [with be thirsty] in the sense "to desire" are generally less frequent in the corpus' (Castro-Chao 2021b: 164). Turning now to the discussion of complementation patterns, Table 7 shows the raw frequencies for each of the documented patterns of be hungry, in correlation with the different senses conveyed by the construction -i.e. literal, metaphorical and overlapping. Percentages are given in brackets and figures are arranged by order of overall frequency. Figure 6 provides the relative frequencies distributed across 50-year subperiods.
Overall, the data unveil a marked preference for one-place complementation (92.8%, 128 tokens), which is illustrated in example (18) above; prepositional complements are found at a much lower frequency, with only 7.2% of occurrences (10 tokens, examples (19) and (20)). This stands in contrast with the data for the verb hunger, which overall prefers prepositional complements (52.8%; see Table 3 and Figure 2).
From a diachronic perspective, the predominance of one-place complementation with be hungry remains fairly stable across subperiods, which is paralleled by a relative low proportion of prepositional patterns. Yet again, this is in contrast with the data for hunger, since this verb experiences notable changes throughout EModE as far as the syntactic   (16) and (17) with hunger, Section 5.1).

Discussion
This section addresses the two research questions raised in Section 1. The first one concerns the possible motivations for the metaphorical extension of the verb hunger from the physical-sensation sense 'to feel hunger' to the emotion sense 'to desire'. The second concerns the reasons why the verb competed with the adjectival construction be hungry in EModE for the expression of the literal sense 'to feel hunger'. In connection with the first question, the analysis of data has shown that hunger progressively abandons the physical-sensation sense in favour of the extended sense 'to desire' (see Table 2 and Figure 1). It has also been explained that this process of   semantic change involved not only a process of semantic transference but also a change in the verb's valency from one argument in the literal physical-sensation sense to two arguments in the extended emotion sense. One possible motivation for this semantic change lies in the religious contexts in which the verb is commonly found. Specifically, the transference may have been facilitated by overlapping examples such as (12) (Section 5.1), repeated below as (21). As already mentioned, the original literal sense in example (21) develops the figurative interpretation 'to hunger for/desire spiritual salvation' because the Needed, which denotes a concrete edible substance, is metaphorically interpreted as 'spiritual salvation' in the context of Christian doctrine. Thus, the meaning of the verb shows properties of both of its two main senses, i.e. 'to feel hunger' and 'to desire' (see Section 5.1).
(21) (=12) Wherfore my deare frendes thys breade let vs specially hungre for, and this drinke let vs thurst for. (Richard Taverner, 1542, On Saynt Andrewes day . . .) In other words, overlapping examples like (21) draw properties both from the original domain of Physical Sensation and from the domain of Emotion. On the one hand, they preserve from Physical Sensation the fact that the noun collocates make reference to a substance which is susceptible of satisfying physical hunger -i.e. breade in (21). On the other hand, overlapping examples draw from the domain of Emotion the fact that the verb takes two arguments instead of one, thereby complying with the bivalent structure of 'to desire', but not with the univalent structure of 'to feel hunger'.
The considerations above lead to the tentative conclusion that overlapping examples facilitated the use of the verb in structures with two expressed arguments, even when the noun collocates were associated with the original domain of Physical Sensation. This phenomenon, therefore, acted as a link between the univalent sense 'to feel hunger' and the bivalent sense 'to desire', and eventually witnessed a semantic transference process which involved the following sequence: 1) The verb denotes the sense 'to feel hunger' (example (11), Section 5.1).
2) The verb denotes the overlapping sense 'to feel hunger for/desire spiritual salvation' (example (21) above).
3) The verb denotes the sense 'to desire something (of whatever nature)' (example (10), Section 5.1). Figure 7 below represents the cline involved in the semantic change of hunger. Robinson (2010: 89) points out that overlapping senses commonly allow 'for creating conceptual extensions and links between senses', which means that they act as a bridge between one meaning category and another. Robinson also points out that in processes of semantic change overlapping senses may (but do not have to) represent the 'first signals of a particular category being used in a novel way ' (2010: 102;also Núñez-Pertejo 2017: 71).
The cline presented above can account for the fact that the overlapping sense is attested mostly in the early subperiods S1 (4 tokens) and S2 (4 tokens) (see Table 2 and Figure 1), but barely so in the late subperiods S3 (1 token) and S4 (0 tokens). It seems that overlapping examples are attested during the 16 th century, while the original sense prevails over the metaphorical one, but once the metaphorical meaning becomes predominant during the 17 th century the overlapping sense fades as the rate of metaphorical use increases. This may have been further prompted by the fact that the adjectival construction expands in the expression of the sense 'to feel hunger', and hence intrudes into the semantic space previously dominated by the verb (Table 5 and Figure 4). Another aspect which needs to be taken into account is the possibility that Latin interference may have played a part, since the verb hunger commonly translates Latin esuriero 'I hunger' in such well known Biblical passages as Blessed are those who hunger and thirst for righteousness (Mathew 5:6; see also Möhlig-Falke 2012: 162). Given that the Latin original already carries a figurative interpretation as 'to desire', Latin influence may indeed have also prompted the semantic transference of the verb from the domain of Physical Sensation to the domain of Emotion. 7 Lastly, it is also noteworthy that metaphorical extension processes like the one observed here tend to involve some sort of perceived similarity between the two senses (see e.g. Nevalainen 1999: 443). As Nevalainen puts it, metaphors 'transfer lexemes from one field of discourse to another on the basis of physical or functional similarity ' (1999: 443). In the case at hand, the similarity between the two senses lies in the perceived lack of something which is missing from the person's current experience. The literal sense 7 The possibility that Latin influence acted as the sole factor triggering the semantic shift of hunger does not seem likely if we take into account that Latin interference can be seen at work already in OE. This is evident from the fact that the OED (s.v. hunger,v. †4.) documents the sense 'to desire' in a text that is dependent on Latin (West Saxon Gospels; see Section 4). Yet the next attestation of this sense is not found until about four centuries later, in the 14 th century (see also MED s.v. hungren v. 1. (c)). In other words, if Latin influence alone were ultimately responsible for the semantic transference process, we would expect it to have happened much earlier in the history of the verb, when the first Biblical translations of Latin esuriero are found.
involves a Feeler that lacks food, and the metaphorical sense involves a general lack of something longed for. Along similar lines, the overlapping sense involves a lack of spiritual union with God, which therefore represents the tertium comparationis between the two domains of Physical Sensation and Emotion.
The second research question addressed concerns the motivations behind the decline of hunger and the parallel rise of the adjectival construction be hungry in the sense 'to feel hunger'. In order to arrive at a semantic characterisation of the verb (e.g. examples (14)-(17), Section 5.1) and the adjectival construction (e.g. examples (18)-(20), Section 5.2), a distinction needs to be made in terms of the force-dynamic schema laid out in Figure 8.
From a force-dynamic perspective, the participants involved in the SoA may represent either an Initiator -i.e. the participant who is responsible for triggering the SoAor an Endpoint -i.e. the participant to which the SoA is directed and which suffers its effects (Croft 1991: 166-167;2012: 197-208;Castro-Chao 2021b). The use of verbs in the active voice commonly (though not always) involves the presence of a controlling Initiator. Therefore, the primary participant in the sentence She worried -i.e. example (22a) below -may be said to be in control of triggering the emotion of worry insofar as she intentionally directs her attention towards the unexpressed object of worry (see Wierzbicka 1999: 302 Conversely, in the case of the adjectival construction She was worried -i.e. example (22b) above -the participant she is not viewed as responsible for triggering the emotion of worry. Rather, she turns out to be merely affected by the internal change of state caused by the emotion itself, which overwhelms her in an uncontrolled manner (see Castro-Chao 2021b). Hence, the participant she represents an affected Endpoint rather than an Initiator.
From the above, it follows that adjectival constructions defocus the Initiator of the SoA, with the dynamic relationship between an Initiator and an Endpoint being left in the background. As a result, adjectival patterns assign a stative interpretation to the SoA, which is not controlled by any of the participants involved in it. By contrast, the active use  of verbs assigns an active interpretation to the SoA, since there is an Initiator which is in control of bringing it about (Wierzbicka 1999: 302).
Taking this into account, the hypothesis is put forward that the loss in frequency of the active use of hunger with respect to the adjectival construction be hungry may be at least partly motivated by the semantic properties of dynamicity and control. On the one hand, the dynamic and controlled use of the verb correlates with the expression of the sense 'to desire' (see Table 6 and Figure 5), which involves an inherently dynamic and controlled SoA (see Section 4). On the other hand, the stative and uncontrolled adjectival construction competes with hunger for the expression of the literal sense 'to feel hunger' (see Table 5 and Figure 4), which in turn correlates with an inherently stative and uncontrolled SoA (see Section 4). On a more general level, this allows us to hypothesise that in EModE hunger -in the sense 'to feel hunger' -loses ground to be hungry as a consequence of the semantic match between the lexical semantics of that sense and the semantics of the adjectival construction itself (cf. Castro-Chao 2021b). In addition, the spread of the adjectival pattern may also be connected to the inherent semantics of impersonal constructions, whose loss took place over the period 1400 and 1500 (see Section 1). As Möhlig-Falke (2012: 149, 193; see also 2017) points out, the first argument in impersonal constructions represents the Endpoint of the SoA, in parallel to the subject of passive patterns. This leads to the tentative hypothesis that adjectival patterns might have come to occupy a functional niche previously occupied by impersonal patterns, since they likewise enable the conceptualisation of the first argument as the Endpoint of the SoA.

Conclusions
This study has considered the syntactic and semantic development of the impersonal verb hunger in the EModE period. From a semantic perspective, a process of semantic change takes place whereby the original sense 'to feel hunger' becomes gradually disfavoured while the sense 'to desire' increases in frequency. This process of semantic change involves a transference from the original domain of Physical Sensation onto the domain of Emotion, and it also brings about a change in the valency of the verb from one argument in the physical-sensation sense to two in the emotion sense.
From a syntactic perspective, the verb features most prominently in patterns with prepositional complements, which convey primarily the sense 'to desire'. From a diachronic perspective, prepositional patterns grow at the expense of one-place complementation, whereas NP and clausal complements remain at a low frequency throughout the entire period. The data further show that the adjectival construction be hungry grows in frequency at the expense of the verb. However, the competition between both modes of expression takes place in the domain of Physical Sensation, and not in that of Emotion. Syntactically, the adjectival construction differs from the verb in that the former predominates in patterns with one-place complementation, rather than with prepositional complements. One-place complementation, in addition, remains prevalent throughout the entire period, whereas prepositional complements remain highly infrequent.
With regard to the research questions formulated in Section 1, a first hypothesis has been put forward as to the motivations which may have led to the semantic extension process of hunger. Given that this verb is commonly found in religious contexts in the corpus data, as well as in the OED and MED entries (see Section 5.1), it seems possible that overlapping examples, which are found primarily in religious texts, facilitated the extension from a physical hunger for food to a spiritual hunger for union with God. This, in turn, led to the expression of a more general kind of desire for something in an abstract, metaphorical sense.
With regard to the decline of hunger in parallel to the rise of the adjectival construction be hungry in the sense 'to feel hunger', the hypothesis is that this may be at least partly motivated by the interaction between lexical semantics and constructional semantics. The sense 'to feel hunger', which is inherently stative and uncontrolled, shows a preference for adjectival constructions which assign a stative and uncontrolled interpretation to the SoA. By the same token, the sense 'to desire', which is inherently dynamic and controlled, is better expressed by the active use of the verb which similarly allows a dynamic and controlled interpretation.
As has been previously mentioned, the EModE period witnesses a clear tendency for the adjectival pattern to rise at the expense of the verb pattern, when in the sense 'to feel hunger'. This process of syntactic change, however, is not completed within the period investigated here -cf. the consistent trend in S4 to reverse the increase/decrease of data in the S1-S3 periods (see Figures 4 and 5). In view of this, the present study might be fruitfully expanded to examine in detail the further course of development of hunger and be hungry during the Late Modern English period. A preliminary analysis of data retrieved from the Corpus of Late Modern English Texts, version 3.0 (CLMET3.0, 1710-1920De Smet, Diller & Tyrkkö 2013) confirms that, as could be expected, in the sense 'to feel hunger' the verb hunger has continued to lose ground, when compared with the EModE data presented in Table 2 above: in CLMET3.0 the relative percentages for the two verbs senses are 70.4% for the sense 'to desire' (38 tokens, cp. 52.3% in Table 2) vs. only 29.6% for the sense 'to feel hunger' (16 tokens, cp. 42.5% in Table 2).
On the other hand, given that similar processes can be observed in the history of English with other impersonal verbs such as shame vs. be ashamed (Möhlig-Falke 2012: 15) and thirst vs. be thirsty (Castro-Chao 2021b), this study may also be extended to those two impersonal verbs, with a view to examining their histories in relation to the adjectival constructions be ashamed and be thirsty. A comparison could also be drawn between these verbs and the data for hunger/be hungry analysed here, as this would allow us to investigate whether the hypotheses put forward in this paper are matched by the histories of other members of the class of desire.