| AUTHOR: Eckardt, Regine
TITLE: Meaning Change in Grammaticalization
SUBTITLE: An Enquiry into Semantic Reanalysis
PUBLISHER: Oxford University Press
Paul Isambert, Université de Paris 3, Sorbonne-Nouvelle, Paris, France.
Regine Eckardt's volume tackles a well-known issue, meaning change, with a
well-known theory too, formal semantics; the novelty lies in the combination of
the two, of course, which to my knowledge has hardly ever been done, and
certainly not to the extent of a book. The bulk of the volume consists in four
case studies; two of them reconsider traditional instances of language change,
namely the development of the English ''going to'' future and of the French
negation; the other two examine less studied facts in German concerning the
history of ''selbst'' and ''lauter''.
The first chapter is a general introduction to the aims and method of the book.
Eckardt subscribes to the view that meaning change is driven by ''pragmatic
inferencing'', i.e. some meaning formerly conveyed as an implicature becomes
conventionalized. This is not a sufficient condition, however, since countless
implicatures are made in any normal conversation, some of them quite repeatedly,
without any meaning change taking place. Besides, some equally promising
constructions develop in some languages and not in others, like the ''going to''
future, which exists in both English and French, but not in German, although the
''gehen zu tun'' (literally ''go to do'') construction exists with the same
''pragmatic potential''. When grammaticalization takes place, according to the
author, speakers ''solve a semantic equation'': some additional meaning (pragmatic
inference being conventionalized) of the sentence cannot be accounted for by the
normal compositional rules and the lexical meaning of words and constructions
and thus must be redistributed to some of its parts, whose meaning therefore
changes. The author advocates the idea that this semantic mechanism may be
better understood with formal semantics, thanks to its explicitness with regard
to composition and the syntax-semantics interface and its attention to
functional elements as well as content words.
Chapter 2, ''Meaning Change under Reanalysis: Previous Views'' reviews some of the
major positions on language change. Eckardt reminds the reader of the classic
contemporary works by Hopper & Traugott (1993) or Lehmann (1982 ) and
replaces reanalysis in these biggest pictures, noticing that not all instances
of grammaticalization rest on reanalysis (although it remains a major force),
whereas reanalysis does not necessarily entail grammaticalization, as discourse
markers illustrate. She then recapitulates three important views on meaning
change: the oldest, the idea that meaning ''bleaches'' in grammaticalization and
also generalizes to broader contexts, has been somewhat tempered by
counterexamples and above all the observation that what is lost in concrete
signification is gained back on the discursive side, i.e. grammaticalization
shows ''pragmatic strengthening''. The second view, held after more general
assumptions of cognitive grammar, sees in metaphor (mapping from one domain to
another, e.g. movement in space to movement in time) the driving force behind
meaning change; however, this view doesn't account for pragmatic strengthening
either, let alone the fact that many changes are not driven by metaphor.
Finally, the author presents the metonymy-based view of historical pragmatics,
whose thesis is that meaning change takes place at the sentence level, when
pragmatic implicatures are conventionalized in utterances where a given
construction appear (and hence is reanalyzed). There remains the issue
characterizing ''onset contexts'', in the author's terminology, i.e. those
contexts where reanalysis was bootstrapped in the first place; expressivity as
well as subjectification has been proposed, but some more light needs to be shed.
Eckardt also reviews some research on discourse markers, which are in a class of
their own in studies on language change. Most importantly, it is far from
obvious that they grammaticalize, since they often are grammatical markers from
the beginning. Besides, they challenge all views concerning syntactic
bondedness, loss of scope, etc., since becoming a discourse marker implies
becoming syntactically freer and gaining scope.
The third chapter, ''Truth Conditional Semantics'', introduces the reader to the
main tenets of formal semantics. The author defends the idea that understanding
a word or a sentence implies being able to say whether it is true or false in a
given situation, thus echoing the basic mechanism of model-theoretic logic (and
semantics). Additionally, ''we use language under the faithful assumption that
our judgements and those of the other will, in the normal case, be in concord.''
So the logical approach is justified, all the more as it is augmented with
possible worlds (to account for, e.g., counterfactuals) and ontological entities
such as points in time or events. The author presents lambda logic and stresses
the importance of comprehension beyond literal meaning operated by logical
entailment, word and world knowledge, and default reasoning. These are crucial
to the computation of Gricean maxims and, since the author defends the pragmatic
view on meaning change, crucial to the rest of the book. Additional tools of
formal semantics are introduced in the following chapters, as required by the
subject under scrutiny.
The next four chapters are the actual case studies. Chapter 4, ''What is Going to
Happen'', reopens the case of the English ''going to'' future. Corpus
investigations show that this construction took off between 1550 and 1650. It
occurred in two structures: with the auxiliary ''be'' (''I am going to visit a
friend'') and in participial clauses (''For Zelmane seeming to strike his head,
and he going to warde it, withall stept backe as he was...'', Philip Sidney,
1593). Comparing literary texts and correspondences, Eckardt suggests that the
construction appeared in written discourses and not oral ones. She also notes
that the addressee of a letter and the audience of a play cannot know whether
any actual movement is involved; this is obvious for letters, while for plays
she argues that the motion of actors may be used simply to clear the stage. The
''going to'' construction offering a convenient way to express one's intentions,
the situation was ripe for a reanalysis.
The semantics of tense is introduced, following Reichenbach (1966) and Kamp &
Reyle (1993), and the ''going to VP'' construction is defined so as to describe a
movement preparing an event (the VP), with the inference that the latter event
will happen soon. This conventional implicature is attributed to ''be going to'',
whose progressive aspect is thus lost. Most importantly, the temporal
information conveyed by this structure distinguishes between the present tense,
which is still contributed by the construction, and an ''IMMINENT'' predicate,
which takes as arguments the reference time and the time of the event, to the
effect that the latter is posterior to, but not too far from, the former.
According to the author, this division of labor explains why the ''going to''
future is acceptable in ''if'' clauses, unlike the ''will'' future. Indeed, it is
really a present tense with some additional temporal information.
Chapter 5, ''From Step to Negation: The Development of French Complex Negation
Patterns'', addresses the well-studied history of French negation. From Latin's
verbal negation with ''non'', Old French added the so-called emphatic particles
based on the nouns ''pas'' (''step''), ''personne'' (''person''), ''rien'' (''thing'', from
Latin ''rem''), ''point'' (''point''), ''goutte'' (''drop'') and ''mie'' (''crumb''), which
ended up as conventional negations (so much so that modern French has virtually
dropped the pre-verbal particle ''ne''), thus following the ''Jespersen cycle''.
However, in Old and Middle French there appeared ''puzzling uses'', where the
particles were used neither in their old concrete meaning nor in their new
abstract one. Eckardt's thesis is that these uses aren't puzzling in any way,
but indeed expected. First of all, they are reported in some contexts only,
namely downward monotone contexts, as illustrated by the following example:
''Öistes vos s'il vendra mie?'' (lit. ''Heard you if-he will come 'mie'?''), ''Did
you hear if he will ever come?'' Consequently, Eckardt considers those particles
negative polarity items (NPIs), i.e. items that occur in ''negative'' contexts
only, like English ''any''.
Eckardt also makes use of Rooth's (1992) theory of focus, to the effect that the
meaning of a focused expression is a set of salient alternatives on which focus
sensible constructions operate; emphatic assertion is such a construction, whose
contribution is that the uttered sentence is more surprising than the
alternatives summoned by the focused constituent. Thus, ''Even Peter knows Lady
Di'' means that attributing that knowledge to Peter is more striking than
attributing it to the set of people evoked by the stress on ''Peter'' (e.g.
Peter's entourage). The final step in the author's theory is to assume that NPIs
always evoke alternatives and that they ''carry the lexical information that they
must always be used in emphatic statements.''
Eckardt then reconstructs the development of French negation as follows: the
above-mentioned nouns were used in emphatic constructions, denoting minimal
quantities and evoking larger alternatives; at some point, this focus meaning
became the basis for a new sense of those words, turning them into NPIs, which
paved the way for emphatic negation, but also for the puzzling uses. Afterwards,
the emphasis is lost.
The next chapter, ''From Intensifier to Focus Particle'', studies the German
''selbst'' couple: the intensifier ''selbst/-self'' (henceforth ''selbst1'') on the
one hand, as in ''Der König selbst öffnete die Tür'' (''The king himself opened the
door''), the focus particle ''selbst/even'' (''selbst2'') on the other, as in ''Selbst
der König verstand den Witz'' (''Even the king understood the joke''). According to
the author, the former, necessarily in focus, evokes alternatives to the
associated noun with centrality effects: the alternative entities are members of
the ''entourage'' of the noun, which is central. Besides, ''selbst1'' is not
necessarily additive, e.g. ''Der König selbst öffnete dit Tür'' does not entail
that anybody else opened the door (contrary to ''selbst2'').
This meaning is the most ancient one, out of which ''selbst2'' derived. Eckardt
shows that ''selbst1'' was used in German poetry in the seventeenth century in
contexts were the only reading was additive and the alternatives could be
computed on the basis of the associated name being in focus and not ''selbst''
itself. Besides, those contexts didn't support centrality effects very well, and
since centrality effects belong to the focus reading of ''selbst1'', they
vanished. This is how ''selbst2'' appeared, whose semantics is equivalent to
English ''even''. This analysis accounts for differences in the syntax, semantics,
and prosody of nowadays ''selbst''.
The last case study, ''To Be or Not to Be a Determiner'', addresses the two
readings of German determiner-like ''lauter'', which may mean ''only'' (with the
additional sense of ''being many''), as in ''Die Maiers haben lauter Töchter'' (''The
Maiers have only daughters, and many of them''), or ''a lot of'', as in ''Unter dem
Baum wachsen lauter Hallimasche'' (''A lot of honey-fungi grow under the tree'').
Though resembling a determiner in many respects, ''lauter'' has some puzzling
properties that don't line up well with determiners: first, it cannot be
stranded like ''viele'' (''many'') in the ''Steinpilze haben wir viele gefunden''
(''Penny bun bolets, we found many''); second, the ''only''-reading violates Keenan
& Stavi's (1986) Universal, according to which quantifiers quantify on the
associated noun and not on the rest of the sentence (''All students attended the
seminar'' requires one to examine the set of students, not the set of people who
attended the seminar); finally, the ''many'' reading disappears under negation and
focus, although it is the most readily accessible elsewhere.
The author shows that the modern reading of ''lauter'' derives from the adjectival
''lauter'' (''pure'') through a stage where it meant ''mere'' (with a pejorative
connotation). She exploits the meaning of ''lauter N'' as ''the set of things X
which are N and such that any subpart x is also N'' (''lauter Töchter'' is the set
of things which are daughters and such that any sub-entity is also a daughter
and not, say, a son). Eckardt argues that the two readings of ''lauter'' differ in
contextual restrictions: ''lauter/only'' specify that the entity under
consideration is maximal, while ''lauter/many'' indicates that this entity is
strikingly big. Additionally, in the formal representation, ''lauter'' is not a
determiner, and ''lauter N'' is more like a bare indefinite which contributes a
property; modern use corroborates this analysis.
In the final chapter, ''Semantic Reanalysis: The Algebraic Backbone of Meaning
Change'', the author proposes a view of reanalysis as a three-step process: in a
pre-stage, there is pragmatic enrichment of a construction, until it is so
loaded pragmatically that the ''cost'' of understanding may lead a hearer to
hypothesize that it is in fact another lexical entry. Actual reanalysis takes
place at the turning point, followed by a post-stage where the new item extends
in use. Eckardt thus argues that semantic reanalysis is a discrete process, with
apparent gradualness as an epiphenomenon that the concept of actualization
(Andersen, 2001) can explain. This process is also as precise as equation
solving in algebra, as illustrated by the ''differentiated patterns'' of use after
The book also contains a small appendix summarizing the basic definitions of
predicate logic augmented with the lambda operator.
First of all, I must say that Regine Eckardt's book is very well-written and a
pleasure to read. Few volumes on formal semantics display such a wealth of
ontological entities, from ''wolpertinger'' to mushrooms, with a nice deal of
German poetry. This is beyond any scientific (dis)agreement.
The volume contains a few typos, some of them worth correcting. Note 4 on page
70 reads (I use ''L'' to mark a Greek lambda for the lambda operator): ''All
examples in the remainder of the book will be set up such that 'variable x to
the left of prefix Lx' and 'variable x bound by Lx' are interchangeable.'' I
suppose the author meant ''variable x to the RIGHT of prefix Lx.''
In note 1, page 93, the author writes: ''Pérez (1990) erroneously claims that
French has a 'venir de faire' future (...). However, (...) there was no such
thing as a French 'venir faire' future (the construction is in fact used to
describe the immediate past).'' I suppose ''venir faire'' stands for ''venir de
faire'', as in the first quotation, or the discussion would be hard to
understand. Besides, ''venir de faire'' is indeed a construction used to express
the immediate past, while ''venir faire'' expresses movement with intention (''to
come here to do something''); it cannot be understood as a future anyway.
Finally, for the reader who might be uneasy with logic, there is an exclamation
mark in the formal representation (4.50c) on page 116 that looks like a real
slip of the hand, as far as I can see.
This volume is an important contribution to the body of work in formal
semantics. Too often are formal semanticists more concerned with nicely carved
representations than with actual data and real attested speech. Donkey sentences
and quantifiers are a very restricted area of language use, and not always the
most interesting. Instead, Eckardt tries to apply the theory to one of the most
important fields in linguistics; this is a crucial move because formal semantics
was never devised to address language change. It is thus here that it can put
its validity to the test. If formal semantics is on the right track, then it
should successfully account for the evolution of languages.
I cannot judge whether the test is passed or not. Like in most attempts in
linguistics, falsifiability is not a decidable matter here. So let's phrase it
otherwise: what does formal semantics add to our understanding of language
change? Is there some new light shed? I'm afraid I must answer negatively.
Eckardt's analyses are fine-grained, careful, and often bold. She does not
refrain from addressing well-studied instances of grammaticalization, offers
precious contributions to old puzzles (like French negation), and at the same
time investigates thornier issues like her two studies on German. But I feel
that this approach would have been equally successful without formal semantics.
In my opinion, the most crucial stances of the book are not couched in formal
terms. Eckardt's defence of pragmatic inferencing, her view of reanalysis as a
discrete and precise process, those do not hinge on formalism in any way.
Besides, the touchstones of her logical analysis, e.g. the introduction of an
''Imminent'' predicate in chapter four, or the contextual specification to the
effect that ''lauter/many'' is used when the object attracts the speaker's
attention in chapter 7, could be described equally well in any other way,
including plain words or the semantic features of structural semantics. I'm
afraid that one may feel, for instance, that since one says that ''selbst1'' is an
identity function whose focus value is a set of functions, one is more
''scientific'' than when one simply says ''selbst1'' contrasts the associated noun
with those salient entities of its entourage. Eckardt argues that Alt(king),
where ''Alt'' is the set of functions mentioned above, yields ''queen,'' ''guard,''
etc. ''All other individuals,'' she writes, ''are linked to the original referent
'a' by being functionally derived from it or, in other words, 'a' is central in
the set of induced alternative individuals.'' Does this mean that being related
by a function has some special effect in the speaker's mind, or that the
function itself has something special (it has been specially devised)? The
former case is totally out of the question, of course, because any two objects
can be linked by a function; and the latter owes nothing to logic, since the
content of that function depends on the linguist, not the formalism.
My objection would be benign if logic was but a descriptive tool. However, the
origin of formal semantics lies mostly in Montague's (1970) faithful assertion:
''There is in my opinion no important theoretical difference between natural
languages and the artificial languages of logicians.'' And Eckardt herself
defends the idea that a speaker's evaluation of a sentence enables her/him to
say in which circumstances it is true or false. But the latter fact, which might
lend itself to endless debate, is just one similarity with model theory. It does
not mean that speakers obey the law of the excluded middle (prototype theory may
say the contrary) or that they have a hard-wired representation of entailment,
let alone compositionality or univocity.
To phrase it differently, the preciseness of formal semantics is of course a
desideratum. This is one important aspect of Regine Eckardt's book; but this
preciseness could be achieved by any sufficiently developed theory, that is a
theory with well-defined and well-articulated concepts. That this theory should
be logic is, on the other hand, a very surprising thesis at best.
To conclude, I would really recommend Regine Eckardt's book for her
investigation of well-known and lesser-known phenomena. My position regarding
formal semantics may seem severe, and if the reader considers my objections
pointless, s/he can run to this volume, which is probably one of the most
interesting in this field. However, since the author was bold enough to venture
logical semantics into the intricate realm of natural language in the wild, I
think this should be taken seriously, hence the tone of this review.
Andersen, Henning (ed.) (2001), _Actualization: Linguistic Change in Progress_,
John Benjamins, Amsterdam.
Hopper, Paul & Elizabeth C. Traugott (1993), _Grammaticalization_, Cambridge
University Press, Cambridge.
Kamp, Hans & Uwe Reyle (1993), _From Discourse to Logic_, Kluwer, Dordrecht.
Keenan, Edward & Jonathan Stavi (1986), ''A Semantic Characterization of Natural
Language Determiners'', _Linguistics and Philosophy_, 9, 253-326.
Lehmann, Christian (1982 ), _Thoughts on Grammaticalization_, Lincom, Munich.
Montague, Richard (1970), ''Universal Grammar'', _Theoria_, 36, 373-398.
Reichenbach, Hans (1966), _Elements of Symbolic Logic_, Free Press, New York.
Rooth, Mats (1992), ''A Theory of Focus Interpretation'', _Natural Language
Semantics_, 1, 75-116.
ABOUT THE REVIEWER
Paul Isambert is a PhD student at the University of Paris 3, France. He's
currently working on grammaticalization and discourse structure, especially
concerning topic shifts and anaphora.