Laurel J. BRINTON and Minoji AKIMOTO eds. (1999).
Collocational and Idiomatic Aspects of Composite Predicates in the History
of English. Amsterdam/Philadelphia, John Benjamins (Studies in Language
Companion Series). xii - 283 pp.
Reviewed by Kate Kearns, University of Canterbury.
The papers in this collection trace the development of three kinds of
complex predicate (CP) in English: V-P verbal complexes or phrasal verbs
(e.g. 'write down', 'dust off', 'think over'), P-N-P prepositional
complexes (e.g. 'on account of', 'with respect to'), and V-N complexes
(e.g. 'do one's duty', 'give an answer', 'make a statement', 'have a look
at'), also well-known as light verb constructions (LVC). Most of the
discussion is devoted to LVCs.
The papers are as follows:
'Chapter 1 is an introduction by the editors.
Chapter 2, 'The Origin of the Composite Predicate in Old English', also by
the editors, surveys LVCs in Old English, chiefly with _(ge)don_ 'do' +
N, _(ge)macian_ 'make' + N, _sellan_ , _giefan_ 'give' + N, _habban_
'have' + N, and _(ge)niman_ , _tacan_ 'take' + N.
Chapter 3, 'Composite Predicates in Middle English' by Meiko Matsumoto,
examines LVCs mainly with _haven_ , _taken_ , _maken_ , _don_ and
_yeven_ ('have', 'take', 'make', 'do', 'give').
Chapter 4, 'Composite Predicates and Phrasal Verbs in _The Paston Letters_
' by Harumi Tanabe, discusses the frequency of LVCs with _do_ , _give_ ,
_have_ , _make_ and _take_ , and V + P phrasal verbs.
Chapter 5 is 'Verbal Phrases and Phrasal Verbs in Early Modern English' by
Risto Hiltunen, surveying LVCs with _do_ , _give_ , _have_ , _make_
and _take_ , with a short section on phrasal verbs.
Chapter 6, 'Collocational and Idiomatic Aspects of Verbs in Early Modern
English' by Merja Kyto, deals with the same LVCs as in the previous chapter.
Chapter 7, 'Collocations and Idioms in Late Modern English' by Minoji
Akimoto, surveys all three kinds of construction in eighteenth and
nineteenth century English.
Chapter 8 is a summary and comment paper, 'A Historical Overview of Complex
Predicate Types' by Elizabeth Closs Traugott.
In addition to describing the developing forms of CP, central themes
through the papers are to examine CPs in the light of general trends found
in language change -- grammaticalization, lexicalization, idiomatization,
and the increase in analyticity in English syntax.
The tendency towards greater analyticity in English syntax, in which
periphrastic expressions take on functions hitherto expressed by lexical
items, is well demonstrated. The editors' introduction reviews the notion
of 'semantic spreading' associated with increased analyticity -- semantic
content which might be bundled into a single word, such as a verb, is
spread across two or more words. The editors cite commentary which
emphasises the displacement of verbal content onto the element following
the verb, leaving the verb as a 'quasi-auxiliary', although the data
indicate that the verb commonly retains part of the predication. In phrasal
verbs the contents of manner of action and of event completion, or
telicity, are carried by the verb and preposition/particle respectively. In
LVCs, although the nature of the event is expressed in the nominal
complement, the light verb may express some thematic content, as in, for
example, the contrast between _maken confessioun_ 'confess one's sins'
and _taken confessioun_ 'hear (another's) confession' (Matsumoto: 70).
Of the CPs considered, complex prepositions fit most easily into the class
of lexicalizations -- an expression like _on account of_ functions
syntactically and semantically very like a simplex preposition. The picture
with collocations like _in love with_ , phrasal verbs and LVCs is less
clear. As Traugott points out (p. 258) collocations which have become
routinized can count as lexicalized strings, and as semantic
unpredictability develops, also show signs of idiomaticization. To some
extent the picture is complicated by varying definitions of what
constitutes a lexical item, collocation, or idiom.
Several authors conclude that the CPs show signs of grammaticalization,
although the discussion of the grounds for this conclusion is fairly brief.
Relevant characteristics of CPs which are covered are the expression of
aspectual meaning in LVCs and phrasal verbs, the decategorialization of N
in LVCs and complex prepositions, and a perceived drift towards more
abstract meaning. Some of these points are taken up below. Another
potentially relevant characteristic is the semantic 'bleaching' of light
The development of complex prepositions such as _in spite of_ might be
considered a kind of grammaticalization, in that the noun _spite_ loses
any overt signs of nominal class -- it cannot be pluralized or modified,
cannot be referential and no longer corresponds in form and meaning to a
full NP. The whole routine is a preposition, and prepositions are a
relatively closed class on the boundary between fully lexical and
functional expressions. On the other hand, unlike other clear instances of
grammaticalization, the development of complex prepositions doesn't
constitute the emergence of a new functional category, but simply enlarges
an existing quasi-functional category.
Where phrasal verbs and LVCs are concerned, I agree with Traugott's
conclusion that grammaticalization is not in evidence, as the new
expressions really must be counted as new kinds of verbs, and no new
functional expressions arise. Although increased abstractness of meaning
and decategorialization of components are common factors in
grammaticalization, they are not sufficient to define its presence.
The evidence for developing idiomaticity discussed in the papers includes
fixed collocation, loss of syntactic freedom, decategorialization of N in
LVCs and complex prepositions, and semantic non-compositionality. As noted
above, idiomaticization overlaps with lexicalization of strings. It is
generally concluded that all three types of CP show idiomaticization.
GENERAL DISCUSSION - VARIATION AMONGST CPS
The strategy of taking CPs as a type of expression, exemplified by phrasal
verbs, complex prepositions and LVCs, and examining CPs for evidence of
general characteristics seems to assume that CPs are in some ways uniform,
and also that phrasal verbs, complex prepositions and LVCs are coherent
classes. Generalizations offered in the volume concerning
grammaticalization and idiomaticization as evidenced by decategorialization
of N, syntactic fixity and restrictions on modification, increasing
abstractness of meaning, and aspectual functions, suggest that the classes
are taken to be fairly internally consistent.
The authors note that there is considerable variation among expressions in
showing these developments, but the general impression of the volume
overall is in line with the editors' summary comment (p. 14) 'The three
structures treated in this volume show signs of having undergone -- to
different degrees and in different ways -- grammaticalization,
lexicalization and idiomaticization.' In keeping with the title of the
volume, there is a greater emphasis on expressions which do show positive
signs of these processes, and less emphasis on an alternative and
conflicting pattern which also seems to emerge in the data and discussion.
DIVERGING DEVELOPMENT PATHS IN CPs ?
There seem to be at least two broad patterns of development. Phrasal verbs
, complex prepositions and some LVCs like _make amends_ move in the
direction of greater idiomaticity, as demonstrated by limited modification,
semantic non-compositionality and syntactic marks. LVCs and complex
prepositions also show decategorialization of N. Syntactic marks include
syntactic limitations, such as restrictions on the passive, and syntactic
characterisitics such as the 'particle shift' construction typical of
Expressions in this group show varying degrees of idiomaticity.
Modification is possible in 'clean this mess right up' but not in 'ring her
right up'; _amends_ in _make amends_ seems fully idiomatic in that
there is no independently occurring noun _amend(s)_ , and yet it can be
modified as in 'make full amends', but the more transparent-seeming _make
friends_ cannot be modified as in #'they made firm friends'. But these
variations don't seem to be signs of different major processes at work --
although there is variation, the general trend seems to be towards more
restriction on modification, consistent with a loss of compositional
Several kinds of LVC stand in strong contrast to this pattern, particularly
in the discussion of modification of N. Matsumoto reviews articles,
adjectives and adverbs in LVCs. Examples (given here in contemporary
spelling for convenience) include 'make great mourning', 'make a foul
affray', 'make a savage attack', 'have great abomination', and so on. Some
expressions are modified adverbially, as in 'make mine avow devoutly' and
'make his complaint piteously'. Matsumoto concludes (p. 92) that
'modification and relativization are the essential features of CPs'.
Tanabe (p. 123) identifies two groups of LVCs: 'To sum up, although most of
the composite predicates are not fixed enough to qualify as idioms, the
examination of nominal modifiers, passivization, verbal and prepositional
substitution revealed that there are some composite predicates which prefer
fixed combination of modification and of verb and preposition: e.g. _do
dever_ , _do part in_ , _do errand to_ , _give grace_ , _have knowledge
of_ , ...'. LVCs unlike this fixed group are described in Tanabe's
general conclusion (p. 130) 'With regard to modification in the composite
predicate, we have found that there is a set of composite predicates which
appear to be relatively free, occurring with a quite restricted set of
modifiers such as zero article, possessive pronouns, _no_, and a limited
class of adjectives. Passivization is possible, but not very common: it
does not affect the high frequency composite predicates which occur in
relatively fixed form'.
Kyto, on LVCs in Early Modern English, writes (p. 199) 'as for idiom
formation, syntactic fixity is evidenced, in particular, in verb + noun
constructions that have zero modification; ... Most uses stand out as
transparent expressions leaving only little room for semantic
Akimoto on Late Modern English provides some of the strongest evidence for
a trend in direct opposition to the decategorialization and
idiomaticization of N in LVCs, discussing anaphoric uses of N and
relativization. Examples include ' something extraordinary will attend his
declaring himself my sister's admirer. THIS DECLARATION will certainly be
MADE in form ..'; 'I have MADE THESE OBVIOUS REMARKS for the better
illustration ...'; 'THE LITTLE ENCOURAGEMENT which was GIVEN by the publick
to his anonymous proposals ..'; and 'THESE HORRIBLE CHARGES that are MADE
against you' (capitals added). Here the NPs are fully referential.
Nevertheless, Akimoto concludes (p. 235) 'The idiomaticization process I
have discussed in this article in respect to the above three structures
involves ... the decategorialization of nouns, ... The phrase containing
such a decategorialized noun eventually becomes an idiom as a whole, and
the undecipherability of its meaning becomes stronger and stronger.'
Akimoto and Brinton's paper on Old English also cites numerous examples of
modification with possessive pronouns, adjectives and some articles -- at
this stage of the language the use of articles was not fully developed, and
so the absence of articles does not have the same significance for the
status of N as in later stages. Akimoto and Brinton comment (p. 51) that
'adjectival modification of the deverbal noun would appear to be one of the
primary motivations for use of the composite predicate, just as it is in
LVCs UNLIKE OTHER CPs
It seems that the importance and frequency of modification makes at least
some LVCs quite different from phrasal verbs and P-N-P complexes. The trend
towards greater analyticity offers rewards with LVCs that are not found to
the same degree with the other CPs. From the point of view of 'semantic
spreading', some 'verbal' content is displaced in phrasal verbs onto the
particle and in LVCs onto the nominal. The particle (or preposition) is a
member of a class which is generally not open to much modification, and
typically modification is absent or confined to intensifiers such as
_right_ . The case with LVCs is very different, because NP structures are
extremely rich, and the range of possible modification is wide,
particularly with a deverbal noun. The choice of a LVC over the
corresponding simple verb may serve purposes of discourse structure, as in
the referential uses noted by Akimoto, and may host a range of
modification, all of which serves to specify the event denoted by the NP.
The role of the NP in a LVC as a host for modification and as a referential
element in discourse may produce a pressure towards greater analyticity and
against decategorialization of N.
LEXICAL COLLOCATIONS AND CONSTRUCTIONS
The notions of grammaticalization and lexicalization as heuristics are in
some ways difficult to apply to such LVCs as in, for example, 'she thanketh
you heartily for the great labour and business that ye have had in that
matter' (spelling modified for convenience) (Tanabe, p. 115), because what
is grammaticalized or lexicalized in the prototypical case is a single
morpheme rather than a phrase, even though syntactic reananlysis is the
context in which grammaticalization is demonstrated. What we seem to see
with LVCs, particularly in contemporary English, is a distinct construction
type. Where the collocation of V and N is routinized, the new construction
can be classed as a lexicalized string, but this does not really express
the fact that the whole NP headed by N is part of the construction. Nor
does it capture the generalization that, for example, 'give the door a
kick' and 'give the rope a tug' seem to be the same construction, and 'make
an inspection of the factory' and 'make a tour of the premises' seem to
both be instances of another construction. Both of these LVC types appear
to be productive, and so a lexeme-based account of them (e.g. in terms of
lexicalization and/or idiomaticization of a lexical collocation) must be
Traugott points out that early stage LVCs have much in common with the
basic constructions of Construction Grammar, such as the two constructions
for _give_ in 'Kim gave Bill a painting' and 'Kim gave a painting to
Bill'. A later stage of development is described as the collocation or
phrasal lexicalization stage, characterized in part by syntactic
constraints. In illustration she cites Goldberg's observation that 'many
metaphorical extensions of _give_ do not allow oblique syntax', as in
*'Kim gave a kick to Bill'.
Traugott does not directly comment on a point that has been of interest in
recent syntactic investigations of LVCs, which is the issue of whether or
not 'a kick' in 'Kim gave Bill a kick' has the same semantic and syntactic
construction role as 'a painting' in 'Kim gave Bill a painting'. In
Construction Grammar the roles are Agent, Patient and Recipient, with
Patient assigned to 'a painting' in both constructions. Traugott's
discussion suggests the same assignment of roles in both basic and LVC
constructions -- she notes (p. 258) that 'CPs have in English become more
and more associated with bare Ns in the Pat role', and mentions (p. 259)
the decategorialization of the Pat NP in CP strings'. The Pat NP referred
to here is the deverbal N. But on the other hand, the simplex verb
near-paraphrase 'Kim kicked Bill' suggests that 'Bill' should be assigned
the Patient role in the LVC as well as in the simple verb sentence. If so,
'a kick' might be a co-predicator, comparable to the particle in a phrasal
verb, or perhaps might be the bearer of a special 'action' or 'eventuality'
role. The latter option would be compatible with the apparent
referentiality of NP in the LVCs noted above, and also with modifiers such
as 'Kim gave Bill a kick', with the near paraphrase 'Kim kicked Bill three
However these unresolved questions on the status of N are resolved, I
suggest that LVCs are probably best viewed as a class of constructions
rather than simply as collocations. A focus on LVCs as constructions
highlights the importance of construction slots, or in other words, the
argument structure of a LVC. As noted above, argument structure is one of
the points of interest in comparing LVCs with the corresponding simple
Authors in the volume have taken somewhat different approaches to elements
in a LVC other than V and N. Three of the seven analysis papers identify
LVCs as of the form V + N or V + N + P, but the underlining of
constructions in examples commonly omits an argument-like PP, as in (p. 38)
the Old English 'sithan he nimth eft lufe to Gode' (afterwards he takes
again love to God) and 'tha niman swythe micle lufe to hyre' (he took such
great love to her), where the _to_ -PP is not marked. Akimoto and Brinton
discuss the frequent occurrence of dative arguments with some LVCs, with
many examples, and note that a dative argument may be alternatively
expressed as a PP. The possibility of dative arguments or PPs in similar
LVCs is shown in the data elsewhere, but generally not commented on -- in
'he criden on seint thomas to yeuen heom milce' (he cried on St Thomas to
give him mercy) _heom_ is not marked, but in the next example 'Bidde we nu
the holigost that he haue milce of us' (bid we now the Holy Ghost that he
have mercy of us) _of_ is marked. (Examples (12b), (12c), p. 65).
Hiltunen cites the examples 'you have given me a knocke on the head' and
'your true Protestant Appetite in Lay-Elder, does a Man's Table Credit',
and comments that 'what may actually come between the verb and the noun may
also depend on the basic structure of the sentence ... Thus, it is not
uncommon to find ... an indirect object in this position.' The comment
suggests that the indorect object in these examples is not considered to be
part of the LVC. Indeed, if the LVC is identified as a collocation then
the arguments are not included, but if the LVC is treated as a construction
then the argument slots are included.
Other kinds of apparent complement to N appear in the data. Foe example,
the Middle English construction _give/have/take/make/do no force_ ,
roughly 'pay no attention to, disregard', seems to require an internal
argument which may be a PP or a bare relative clause. PP arguments are
marked in examples like 'Of religioun thai had na fors' (of religion they
had no force) (p. 64), but the clausal complement is not marked in 'I yeue
no force what felle on me' (p. 62).
Generally, discussion in the volume gives little attention to argument
structure. Along with further investigation into the areas mapped out in
the collection, research into the development of the argument structure of
LVCs would be very valuable, especially given the wealth of data collected
ABSTRACT AND CONCRETE (PHYSICAL ACTION) MEANING
Another pattern which seems to be present in the data and might be worth
further exploration is a general shift in the LVC inventory towards more
concrete, physical action predicates. It is noticeable in the Old English
data that most of the predicates are quite vague, such as _do justice_,
_do harm_ , _do honour_ , _do service_ , etc. A handy rule of thumb for
the action content of predicates is to ask whether or not one could attempt
to mime the content. For these predicates the answer is 'no'.
With a few exceptions (e.g. 'give a cry' and 'take a bath' in Middle
English) the familiar contemporary kinds of LVCs for vocalizations or
noises ('give a snort' , 'give a shout' ) and physical actions ('make a
dash' , 'take a walk', 'have a look at' , 'give X a shove' , 'take a poke
at' ) are largely absent until quite late. Hiltunen writing on Early
Modern English points out (p. 148) that 'collocations of the most concrete
kind, which are so often given as the most typical example of verbal
phrase, e.g. 'to have a look/bath/drink/shave/walk' and the like, are
almost completely absent in the present material'.
The great increase in physical action LVCs in Modern English may be a
reflex of a late stage of development of LVCs, where new LVCs are coined
straight from the simple verb by nominalizing the verb, generally by zero
derivation. Such a source seems particularly plausible for the nominal in
'give the room an air' or 'Can I have a use of your pen?', where the
deverbal noun seems native to the LVC. (In some varieties of English LVCs
like these can only be formed with the gerund.) In particular, _use_ in
'have a use of' is pronounced with a voiced fricative, like the
corresponding verb, and in contrast to the deverbal noun _use_ in 'He had
the use of a car at the time', which has a voiceless fricative. Deverbal
nouns such as these appear not to be fully nominal (unlike 'a kick'),
perhaps not because they have been decategorialized, but because they have
never been fully nominalized. An increase in coining LVCs direct from
verbs would be consistent with an increase in LVCs with basic, prototypical
verbal meanings. This is a significant departure from what appeared to be
a motivation for LVCs in the Old English period, which was to provide vague
predicates, often based on nominals denoting properties of behaviour,
rather than actions (e.g. justice, honour, service, etc.).
THE INDEFINITE ARTICLE AND ASPECT
Several authors comment on the role of the indefinite article in LVCs.
Akimoto claims (p. 225-6) that the progression from forms like 'I might
take an advantage of you' to the modern variant 'take advantage of' show a
move from concrete to abstract meaning. A similar point is made in passing
by Tanabe (p. 118), citing a suggestion that the article, in conveying
individuation, makes an abstract noun less abstract.
More generally, the individuation expressed by the article is claimed to
convey brevity or completeness of action -- several authors make this
point, including the editors' introduction and Traugott's summary paper,
claiming that this aspectual meaning is telicity, comparable to the telic
effect of phrasal verb particles.
Traugott raises an interesting point (p. 246-7) concerning the so-called
imperfective paradox, observing that 'she was making a joke when the lights
went out' does not entail that she joked, but 'She was joking when the
lights went out' does have the entailment. This effect seems to be
variable -- if I was having a look at the painting when the lights went
out, it seems that nevertheless I did have a look at the painting and I did
look at the painting, but some incomplete effect is found with LVCs that
denote blocks of activity like 'give the floor a sweep'. If I was giving
the floor a sweep when I was called away, to some extent I gave the floor a
sweep, but I didn't finish giving the floor a sweep. On the other hand,
'She gave the floor a sweep for five minutes ' seems to be slightly more
natural than 'She gave the floor a sweep in five minutes', but 'She gave
the room a tidy in five minutes' is fairly natural. This may be related to
the fact that _tidy_ lexicalizes endstate but _sweep_ lexicalizes manner
Some LVCs seem to have an atelic effect. For example, 'read the paper' can
be either telic or atelic, and so can be modified by 'for an hour' or 'in
an hour', but 'have a read of the paper' can only be atelic -- *'I had a
read of the paper in an hour', compared with 'I had a read of the paper for
an hour.' Another effect with _give_ + action-N LVCs is the oddness of
goal phrases (which express telicity), as in 'give the rope a pull off the
hook', compared with 'pull the rope off the hook'. Similarly, one can have
a walk around the park or take a walk to the corner, but not have a walk to
the corner. The 'aimless, objectless, individual activity' noted by
Wierzbicka (1982) for 'have a N' seems to clash with classic telicity,
although the LVC may convey brevity and termination of the event. Perhaps
the 'countability' of the event expressed by an LVC is boundedness but not
telicity, according to the distinction made by Depraetere (19).
Complex predicates, and particularly LVCs, vary a great deal in the
properties examined in the collection. Within the class of LVCs there are
many distinct types, and even within what appears to be a single
construction such as 'give X a kick/air/sweep/tidy/pull' there are
differences. In fact LVCs are such an unruly mob that it is extremely
difficult to make any generalization stick, beyond the baldest
characterization of a V + N collocation as predicate. So a strength of the
volume, which is the broad scope of its coverage, also limits what can be
achieved in analysis and discussion, because of the difficulties in
treating as a class a range of collocations so lacking in uniformity.
In the description of the development of LVCs the collection makes a very
significant contribution. Although the presence of LVCs has been noted and
briefly discussed in earlier works, they have not been studied in any
depth. Tracking through from Old English to Early Modern English, the
papers here present a great wealth of text examples and lists of
collocations, drawing on a wide range of sources. This collection is
presented as a preliminary survey of a major area, and it lays a solid
foundation for much further research. It is to be hoped that the research
here will be followed up in detail, building on the advances made in the
The volume will be of great value to anyone interested in idioms and
idiomaticization processes, collocations and constructions, and light verbs.
(I add a final comment to the publishers -- word-by-word glosses on
non-English data are helpful to the reader who may not be familiar with the
language in question. Old English is sufficiently different from Modern
English to count as a separate language, especially for readers who are not
native English speakers. I suggest that glosses on Old English data would
add to the value and interest of historical works like this.)
Depraetere, Ilsa (1995) 'on the necessity of distinguishing between
(un)boundedness and (atelicity.' Linguistics and Philosophy 18, 1-19.
Wierzbicka, Anna (1982) 'Why can you _have a drink_ when you can't
*_have an eat_ ?' Language 58, 753-99.
Kate Kearns is Senior Lecturer in Linguistics at the University of
Canterbury, New Zealand. Her research interests include light verbs,
aktionsarten and the argument structure of verbal predicates in English.
(Dr) Kate Kearns
University of Canterbury
Private Bag 4800