Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.

Semantic Reanalysis and Language Change

Regine Eckardt*
University of Göttingen

Abstract
Reanalysis is a well-known process of language change in morpho-syntax. However, the semantic
composition of sentence meanings can also undergo reanalysis and lead to meaning changes for
parts of the sentence. The article provides the basic notions of compositional semantics ⁄ pragmatics
that underlie semantic reanalysis, surveys possible constellations and causes of reanalysis, and con-
trasts the process to other ways of semantic change. I will, moreover, illustrate semantic reanalysis
on basis of a case study which highlights some of its typical features.

1. Meaning Change and Grammaticalization

Many people enjoy the fascinating task of tracing the history of words and expressions.
Meaning changes often allow insights in cultural history, and major processes of change
can be observed both in history and in daily life. For instance, speakers can decide to
restrict the use of a word to what was formerly just the most frequent referents, like in
Old English mete (‘food’) which turned into modern meat (food like in ‘steak’). This is
called semantic narrowing, and we will use the example to introduce a few convenient
notations.
I will use [[a]] to refer to the meaning of a word a, e.g. [[dog]] stands for the meaning
of the word dog. If the meaning of a word changes, we will use indices to indicate the
two different meanings. Our instance of narrowing looks as follows: [[mete]]old was the
concept ‘food’ or ‘meal’ in general, whereas [[mete]]new is the narrower concept ‘meat,
edible muscle matter of animals’. If we consider more than one shift in meaning, we will
use numbers as indices. For instance, we will later discuss three subsequent meanings of
German fast, notated as [[fast]]1, [[fast]]2 and [[fast]]3. Other wellknown meaning changes
include generalization, like for instance when [[dog]]old = ‘special breed of dogs’ was
extended to [[dog]]new = ‘species canis’. Furthermore, we know metaphoric extensions,
like [[mouse]]old = ‘small rodent’ to [[mouse]]new = ‘computer control’, and also metonym-
ic extensions, for instance when [[press]]old = ‘machine for printing’ generates the newer
sense [[press]]new = ‘newspapers’.
Other cases of language change are fascinating in other ways. When content words
turn into grammatical words or morphemes, it is much less clear how speakers arrive at
meanings as abstract as those of tense morphemes, auxiliaries, evidential markers or quan-
tifiers. What exactly happens when a phrase he is at hunting changes its message from
[[Tom is at hunting]]old = ‘Tom takes part in a hunting event’ to [[Tom is a-hunting]]new =
‘at the time of reference, Tom hunts’? How does the overall shift in the message project
back to the parts, in particular to the interaction of be and –ing, to yield a progressive
aspect marker? Changes like this have been studied as grammaticalization of words or com-
plex expressions (Hopper and Traugott 2003). The ensuing changes in morphosyntax are
usually clear and well-understood. What happens at the semantic side, however, often

ª 2011 The Author

Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
34 Regine Eckardt

cannot be classed easily under the four types above. Another fascinating puzzle is posed
by isolated language changes which fit neither of the common types. For instance, how
did it happen that premis(s)es, known to the logician as ‘the assumptions in a deductive
argument’, in English also means ‘house, estate’? The two concepts are not related meta-
phorically or metonymically, and the two senses don’t generalize or narrow down each
other, either. The present article will describe how semantic reanalysis as a mechanism of
language change can help to model and understand semantic changes like these.
As later examples will suggest, semantic reanalysis can in principle affect any part of a
sentence, and may or may not come along with morphosyntactic restructuring. However,
many instances of language change which have been discussed in the literature, and
where this process plays a role, are instances of grammaticalization. I will therefore briefly
introduce some core concepts of the field without aiming at a full survey.1 Reviewing
many examples of grammaticalization, scholars discovered so-called clines or pathways of
grammaticalization. These are ordered series of denotations a1, …, an such that several
word forms w, w’, w’’ have been found which exhibit these meanings one after another:
At some time t, [[w]]old = ai and at a later stage t’, there is a use [[w]]new = ai+1 in addi-
tion to [[w]]old. The older use may get lost over time, but often also persists. Some words
can travel the full series a1 to an but others might only cover a subseries. What is impor-
tant is that clines are almost always passed unidirectionally, which means that there is no
known word v that showed up with [[v]]old = ai+1 before it developed [[v]]new = ai. Clines
are viewed as typological universals, which means that there could be, and often actually
are, words from more than one language that developed meanings along the cline.2
Collections of clines (like Heine and Kuteva 2002) help to catalogue regularities in
grammaticalization even though they do not by themselves offer an explanation why the
respective semantic positions show up in a specific order. Clines can also illustrate in what
sense grammaticalization is a non-volitional type of language change. If some language,
say, English, contains a word w’’ in a sense [[w’’]] = ai on a known cline, this does not
mean that English speakers can volitionally start to use w’’ in the next sense [[w’’]] =
ai+1. For instance, English possesses the verb become. This verb shares its meaning with
Old German werden which developed a use as a future modal. So, English become and
German werden share a position on a cline. Nevertheless, English speakers could never use
a sentence Italy becomes (to) win in the same futurate sense as its German word-by-word
translation Italien wird gewinnen. Data suggest that semantic changes on clines often rest on
a typical kind of sentence context in which the crucial word was used. German future
werden, for instance, most likely developed in sentence contexts like ‘He became (a) killed
(man)’, under the new interpretation that the participle (‘killed’) is no longer interpreted as
part of an object NP but as part of a verbal complex with futurate meaning (Behaghel,
1924). While the syntactic reorganizations are usually quite clear, the semantic shifts
eluded analysis for more than 100 years. Only since the advent of detailed compositional
semantic analyses (prominently in the framework of truth conditional semantics) was it
possible to trace and spell out the semantic reorganization of constructions (Eckardt 2006,
2003). I will first offer an abstract definition of the process and we will then turn to
examples.

2. Semantic Reanalysis
We follow Frege’s classical assumption that language is compositional: The meaning of
larger phrases and sentences is composed systematically from the meaning of its syntactic
constituents. If a sentence S is hence built up from parts x1, x2, …, xk, then its meaning

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
 Semantic Reanalysis and Language Change 35

[[S]] is construed by combining the meanings of these parts in the intended order: [[S]] =
[[x1]] ¯ … ¯ [[xk]]. Here, ¯ stands for semantic composition in general. If you are
acquainted with truth conditional semantics, you can mostly read ¯ as ‘functional com-
position’. The parts x1, … xk comprise all meaningful parts of S, be it words, mor-
phemes, or sometimes also tacit semantic operators. For instance, the sentence John ate the
beans has the parts John, eat, PAST, the and beans. Syntax will drive the order of combina-
tion, hence the is the determiner to beans, the beans is the object of eat and John its subject,
etc. Assume now that S is understood in a new sense [[S]]new. Assume that the parts of S
are still x1 to xk. This forces the hearer to assume that at least one of these parts xj must
also have a new meaning [[xj]]new or else, the speaker meant to convey her message
[[S]]new in a non-compositional manner.3 At this point, xj changes its meaning due to
semantic reanalysis. Generally, semantic reanalysis follows the maxime ‘guess meaning by
subtraction’. The overall message [[S]] is known. The meanings of almost all parts of S,
x1, … xk except one xj are also known, and the hearer also has an idea about the syn-
tactic structure of S, which determines the order of semantic composition. The hearer
must answer the following question:

(*) [[S]] = [[x1]] ¯ … ¯ [[xj-1]] ¯ <X> ¯ [[xj+1]] … ¯ [[xk]]

What semantic value <X> will make this equation work?

Whenever a word, morpheme or construction xj is used in a new meaning [[xj]]new in

answer to a question like (*), I’ll say that xj adopts a new meaning due to semantic
reanalysis. Note that this definition is intended in a wide sense. The question (*) can arise
because the hearer understands a new message for S, or because he interprets on basis of
a new syntactic structure, or reassignment of pragmatic values. Let us see some examples.
In the simplest case, an old message is restructured in a new way. The NP the premises
as a way to refer to houses and buildings arose in this manner. According to the Oxford
English Dictionary (OED), the noun premise originally was anaphoric: ‘The matters or
things stated or mentioned previously; the aforesaid, the aforementioned, the foregoing.
Often (…) in legal phraseology’ (OED, premise 2.a.). In stereotypical advertising contexts,
this anaphoric element was used to refer to houses: ‘In pl. The subject of a conveyance
or bequest, as specified in the opening part of a deed and referred to later in the docu-
ment; the houses, lands, or tenements previously specified in a deed or conveyance.

(1) All which maners, londs, and tenements, and other the premisses, we late purchased’. (1508
in Nichols Royal Wills 379, OED 3.a.)

In semantic reanalysis, the anaphoric status of the NP was given up in favor of a context-
independent definite expression for larger units of housing (OED, premise 3.b).
Reanalysis happened on basis of examples like (1508): … we purchased the premis(s)es, and
the minimal new assumption of the hearer was that premise was an ordinary noun, rather
than a high register anaphoric element.

[[the premisses]]old = anaphoric element, instantiated in context by the ‘estate’

[[the premisses]]new = definite NP, denoting a given estate

In other cases, morphemes can get grouped in new ways in morphosyntax. The German
preposition während (‘during’) goes back to a present participle of the verb währen (‘last’,
‘continue’) in constructions where it shows agreement with a masculine genitive noun:

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
36 Regine Eckardt

währendesParticiple KriegesGEN > währendpreposition desdet KriegesGEN (‘lasting war’ > ‘during the
war’). Again, the overall meaning of the new-born PP was clear, leading the hearer to
assign a new prepositional meaning (= ‘during’) to the new-born word während.

½½w€ahrend½des Krieges""" ¼ <X> $ ð½½des"" $ ½½Krieges""Þ

In grammaticalization, hearers typically understand a new sentence meaning that differs

from the original [[S]] in subtle ways which force semantic reanalysis. When hearers took
sentences like [[She is going to be married]] as a report about future events, rather than
about present movements, this invited them to also assume that going (to) contributed to
the sentence meaning in a new way (Hopper and Traugott 2003; Eckardt 2006, ch.4).

[[She is going to be married ]]new

= [[She]] ¯ [[be]] ¯ [[PRESENT]] ¯ (<X> ¯ [[to be married ]])
<X> = [[going]]new

In this case, it was most likely the meaning shift that forced a reorganization in syntax
from biclausal going + [to do z]S-inf to the monoclausal going-to do z construction. The
emergence of prepositional während, in contrast, arguably was caused when hearers
replaced a pretty baroque, archaic pattern in German, ‘free genitive circumstantial modi-
fier’, by the more common PP pattern. Here, syntactic analogy was a driving factor. The
premise case, finally, caused no syntactic change at all. Our range of examples shows that
semantic reanalysis is an autonomous mode of change. It can cause, or be caused by,
changes in morphosyntax, and sometimes happens all on its own.
As can be seen from these examples, semantic reanalysis necessarily takes place in a
sentential context, though not necessarily only in cases of grammaticalization. Scholars
have always been curious about what kind of context is endangered by semantic reanaly-
sis. Heine (2002:86) refers to such contexts as bridging contexts, and suggests that meta-
phoric interpretations could be the driving force. However, the semantic shifts in
grammaticalization are rarely of a kind that one would naturally call metaphors. Particu-
larly, bridging contexts are typically not instances of innovative creative language use (as
would be typical for metaphor) but uses that are simply ambiguous between an older
and a newer interpretation – with the newer structure only unambiguously visible
somewhat later in the data. Diewald (2002) addresses them as critical contexts. When we
look into historical data, an utterance S is critical for an expression xj primarily if S, in
the given text passage, makes sense both on basis of [[xj]]old and [[xj]]new. Note that xj
can also change its morphosyntactic category in the reanalysis. Diewald (2006) proposes
that critical contexts are in fact constructions in the sense of construction grammar, i.e.
pairs of an expression and a meaning <e,m> where m is attributed to e without further
semantic composition. Only later, she suggests, will speakers attempt to decompose e
and assign parts of m to the subexpressions of e. We see that this has happened when
we encounter quotes which only make sense on basis of [[xj]]new. They prove that the
change has taken place and are called isolating contexts by Diewald. I do not want to take
a stand as to whether such intermediate idiomatic uses can or can not happen. Some
developments, as Diewald’s case of modal verbs in German, may show a data record
which suggests an intermediate idiomatic use of certain modal constructions. However,
other data records do not show similar phases, suggesting that semantic reanalysis can
also occur on basis of more varied ambiguous usages. We will see an example later in
the paper.

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
 Semantic Reanalysis and Language Change 37

If you observe your own communicative activities, you will find it hard to locate
sentences you read or hear where you’d have the impression that they are ambiguous (in
the structural ⁄ semantic sense that is at issue here) or that you don’t really understand
them. Nevertheless, we can safely assume that the critical contexts for tomorrow’s lan-
guage changes are around us today. So we need to tackle the question what causes a
hearer ⁄ reader to assume that the speaker didn’t intend to use sentence S in the conserva-
tive sense but in a new one? Eckardt (2006, 2009) suggests that sentences S, intended in
conservative language use, can carry a pragmatic overload. To put it simply, S in the
conservative sense and structure would be so hard to interpret and make sense of that the
hearer feels justified to believe that the speaker meant to use S in some new sense. Such
uses of S are called precarious uses in Eckardt (2009), and we will see examples later in the
article. ‘Precarious’ means that the utterance was not really wrong, but it was risky way
for the speaker to convey her message.4 Precarious contexts may in part overlap with
Diewald’s critical contexts, but the term is intended to suggest that the utterance
stretched the limits of the conservative grammar system. (Another differenc is that precar-
ious contexts can spur changes beyond grammaticalization, whereas Diewald appears to
restrict attention to grammaticalization.) Pragmatic enrichment has been suggested as
another process that leads from old literal to new literal content (Traugott 1988; Traugott
and Dasher 2002), but given that pragmatic enrichments are standardly available anyway,
one may wonder why pragmatic enrichment should lead to language change rather than
to a semantically flexible but overall stable system.
In summary, semantic reanalysis offers a viable explanation for semantic change in sen-
tential contexts, rare directions of change and infrequent or singular patterns. It typcially
can’t be actively reinforced, and in this differs from metaphor or metonymy. Scholars and
case studies suggest that semantic reanalysis can

• be triggered as a by-product of structural reanalysis which, in turn, may be indepen-

dently motivated (e.g. less complexity, Lightfoot 1991, modifier-head reanalysis, van
Gelderen 2008)
• be spurred by analogy (Fischer 2007)
• be triggered by the urge to avoid pragmatic overload (Eckardt 2009)
• be triggered by the urge to conventionalize pragmatic enrichment (Traugott and
Dasher 2002).

In the next section, we will look into one case study in more detail.

3. ‘Almost’ by Reanalysis
Words with the denotation ‘almost’ are often derived metaphorically from words that
express spacial closeness (G bei-nahe, E nearly, F prèsque). However, other sources have
been pointed out, including changes ‘badly’ > ‘almost’ (Amaral, 2006), and ‘similarity’ >
‘almost’ (Kytö and Romaine 2005). I will use two German examples here, schier and fast,
to illustrate semantic reanalysis, specifically the non-productive character of reanalysis, the con-
text-dependency of reanalysis, and the pattern of precarious uses that antedate and accompany
the time of reanalysis.
In Southern German varieties, the adverb schier can be used to express ‘almost’:

(2) Der ist schier gestorben vor Angst

‘He almost died for fear’

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
38 Regine Eckardt

The word can be traced back to a temporal adverb schier that meant ‘soon’. Probably
contexts of change are temporal constructions where at a given reference time, some
future state of affairs ‘will soon be reached’ = ‘has almost been reached’.

(3) da nam Thobias von der gallen des visches, er salbet die
there took Tobias of the gall of-the fish he anointed the
augen seines vaters,
eyes of-his father
und er geduldet es schier ein halbe stund.
and he bore it soon a half hour
Bible of 1483, (Tobias 11, 13)

Interestingly, Luther’s bible (1534) uses fast in the same passage. Grimm’s Deutsches
Wörterbuch (DW) hypothesizes that the new sense schiernew arises from uses like the
following.

(4) Als wir schier an 3 Stunden getaget hatten...

when we soon by 3 hours conferenced had

Let us see how two different syntactic-semantic analyses of (4) can be given on basis of
the word’s older and newer sense. Both propositions state that at the reference time, a
conference of nearly 3 hours length had taken place.

½½S""old ¼ ½als schier ½wir hatten an 3 Stunden getaget""

In the semantic composition, the word schier relates a proposition p to a time t.

SOON(t, p) is true iff p is false at t but becomes true at some time soon after t. In the
given example

[[S]]old = SOON(R, p)
where R is the reference time, p is the proposition ‘we met for 3 hours’5

However, a different way of making the same point is this: ‘when we had met for nearly
3 hours’. And maybe, this is even a more down-to-earth way of talking, compared to
statements about propositions that would soon be true. Understood in this sense, the
hearer faces the following semantic composition:

(#) [[S]]new = p is true at R

where R is the reference time, p is the proposition ‘we met for almost 3 hours’6

She would hypothesize the following compositional structure:

½als½wir hatten½schier an 3 Stunden" getaget""

and find that [[S]]new can indeed be composed from the old parts [[als]]old, [[wir]]old etc.
apart from one step: There is a superfluous word schier, and a necessary semantic operator
<X> = ALMOST that is needed to reach (#). I assume a polymorphic analysis of ‘almost’
which can combine with arguments of different logical types (Penka 2006, Eckardt 2007).
The most natural idea for the hearer will be to assume that [[schier]]new = ALMOST. I’d like

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
 Semantic Reanalysis and Language Change 39

to stress that the two denotations [[schier]]old and [[schier]]new are truly distinct, as you can
easily test by comparing, e.g. English soon and almost, or German bald and schier.
Let us now turn to another case, the development of Modern German fast ‘almost’.
Like its English cognate fast (‘quick’), it derives from an older adjective ⁄ adverb fast with
the sense ‘solid, fixed, firm’. The oldest quotes show vast ⁄ fast to express physical firmness
but the word was metaphorically extended to psychological determination, firm pursuit
of a task. From there, it was generalized to intenseness, and high degree. Around 1400,
uses of fast often translate to English very much, determinedly, energetically. In ModHG, fast
has completely lost this sense. The variant fest is used to express physical firmness.

[[fast]]1 = ‘physically firm, fixed’ (ModHG fest)

[[fast]]2 = ‘very much, intense’ (obs.)
[[fast]]3 = ‘almost’

Older and newer denotations are almost contradictory. If a thing is ‘firmly attached to
something’, it is simply not ‘almost attached’, and likewise if a person does something ‘inten-
sely’, that person simply doesn’t do the thing just ‘almost’. Philologists occasionally take
such examples as evidence for a general weak-mindedness of ordinary speakers, but I do not
think that this diagnosis does justice to the human language capacity. I want to argue that
the modern meaning [[fast]]3 arose by reanalysis at a time where precarious uses of [[fast]]2
increased. In the following, I will use fastold and fastnew to refer to these two uses.
The authors of the DW state that ‘isolated uses (of fast) in 16 ⁄ 17th century slip towards
[the ‘almost’] sense’ without speculating about the kind of context or reason why these
slips occur. The most concise claim in this direction is made with respect to a quote fast
um zwo uren ((we arrived) at almost 2 o’clock), and the DW states that the author
conveyed ‘ungefähr, gerade um zwei uhr, weil hier die bestimmtheit leicht an unbe-
stimmtheit grenzt’.7 I find such statements dissatisfying, because I can not see how exact-
ness can overlap conceptually with vagueness. More objectively, perhaps, the quoted
studies on sources for words that mean ‘almost’ do not list a second instance of this seman-
tic shift whereas sources like ‘similarity’ or ‘temporal closeness’ reappear more frequently.
What is in fact typical about this quote is that vast ⁄ fastold ‘very much’ would have required
a gradable argument to combine with, whereas a time point like zwo uren can not be graded.
(In terms of Diewald, the quote shows an isolating context, not a critical context.) Other
non-gradable arguments that occur early with fast include the quantifier alle where again,
‘almost all’ is a reasonable semantic combination whereas ‘very much all’ would be semanti-
cally defective (Kennedy and McNally 2005). We can hence use vast alle as indicators of
early uses of fastnew. Using the source text section of the DW as a search base, I looked up
the collocation vast alle to verify the dictionary’s estimate for early occurrences. (The older
spelling variant vast was given up after 1800; it avoids hits in ModHG). The search yields 13
hits of which 12 date after 1500. An early quote from 1314 remained isolated in all later
searches and I do not take it as an indication for a much earlier language change. A fuller
search of the DW sources for vast yields 797 hits altogether, including a few double hits.
The manually surveyed hits 1–200 reveal that those usages of vast that are dubitable, ambig-
uous, or clearly ‘almost’ uses all date after 1500. The time boundary 1500 is confirmed by
open Google search for ‘vast alle’ (19.5.2010). An evaluation of first 100 hits yields 33 uses
in Early ModHG texts; all after 1500, and a substantial portion even much later. Based on
this evidence, we can preliminarily conclude that fastnew appeared after 1500.
This is corrobated by the Bonner Frühneuhochdeutsch-Korpus, specifically the
sources before 1500. These early sources contain 136 uses of fast ⁄ vast, among which only

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
40 Regine Eckardt

3 are unclear uses, and zero uses in the sense ‘almost’, which confirms that fastnew was
not in use before 1500. The spelling situation looks as follows: There are 87 hits of
the form vast(e), 70 of the form fast(e), but the latter include 60 hits in one document
alone, hence idiosyncratic to its author. Excluding these, we have a 87:10 ratio
between vast:fast spellings which indicates that vast was common if not predominant at
the critical time.
While these searches showed that fastnew arose around 1500, they offered no clue for
the source construction. We’ll therefore turn to Flugschriften gegen die Reformation,
1518–1524 (FgR, Laube and Weiss (eds.), pamphlets that were addressed to the educated
public and argued against Luther’s reformation). In a total of 47 uses of fast ⁄ vast in these
pamphlets, 2 clear uses show the new ‘almost’ sense, 2 uses are ambiguous between ‘very’
and ‘almost’, 9 uses are precarious. All quotes date from 1518 to 1524 and were written
by preachers. The pamphlets show a mix of conservative, innovative, and precarious uses,
quite typical when language change is under way. Let us first look at two clear uses of
fastnew:
(5) … dermassen es nun fast Quinta essentia worden ist … (FgR p. 90)
such-that it now almost quinta essentia become is
‘such that it is by now almost quintessential’

(6) … die selbige seint yrer handwergk halben fast alle zu

the same are (for) their craft’s cause almost all to
betlern worden …
beggars become
‘these have almost all turned to beggars, for their craft’s sake’
(FgR p. 662)8

In both cases, fast plausibly takes an argument that is not gradable (‘become quintessen-
tial’, ‘all become beggars’). Such an argument couldn’t combine with the intensifying
adverb fastold ‘determinedly, very’. The second passage moreover continues talking about
other people who did not turn beggars but earn money with sinful occupations instead.
This contrast likewise suggests an ‘almost all’ interpretation. This shows that two authors
had adopted fastnew in their mental lexicon.
More frequent were uses of fastold in precarious constructions. The following passages
contain the collocation fast... alle but are intended as conservative uses. A reanalysis in the
‘almost’ sense could be possible in both cases. However, the overall text passages do not
suggest that the authors were talking about ‘almost, but not all N’.

(7) Ain gegenwurff, darauff die feind der kirchen fast all
a counter-thesis on-which the enemies of-the church fast all
ir grund setzen.
their reasons put
‘a counter-thesis on which fast all enemies of the church build their reasoning’
(FgR, p. 734)

In a conservative interpretation, the author used fastold ‘fixedly’ as a modifier of setzen

(‘put fast’). However, the modifier is separated from the verb by a definite NP. While
adverb positions are very flexible in Early ModHG, definites tend to be placed before
manner adverbs. Hence, syntax suggests an alternative structure in which fast modifies the

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
 Semantic Reanalysis and Language Change 41

NP all ir grund. This excludes fastold for semantic reasons, and can at least raise doubts in
the reader as to what the author intended to say.

(8) (one should not say anything against selling indulgences)

… welches doch fast in allen artickeln geschieht.
which PRT fast in all articles happens
‘which however happens fast in all articles’
(FgR, p. 61)

If we look for an argument for fastold, the verb ‘happen’ seems the only potential candi-
date. What the author seems to have had in mind is a combination of fastold with the
antecedent of anaphoric geschieht: ‘in all articles, people argue determinedly against selling
indulgences’. However, this reading is only accessible after anaphor resolution; moreover
the modifier fast is again separated from its argument. Syntax suggests a combination (fast
+ ‘in all papers’). This can only be interpreted in terms of fastnew. However, I suspect that
the writer once more simply put things in a precarious way, as the surrounding text does
not suggest an ‘almost’ reading.
While the previous examples show precarious uses of adverb plus verb, other quotes
show that a propositional use was established at the time.

(9) und das ist alles gescheen durch das vorgissen des bluts
and that is all happened by the shedding of-the blood
Christi, wan keyne vorgebunge der sunde mag (…) ane
Christ’s if no forgiving of sin may without
blutvorgissunge bescheen, ßo fast alle ding durch das blut gereyniget
blood-shedding happen so fast all things by the blood cleansed
werden.
become
‘and all this happened by the shedding of Christ’s blood, as much as there is no
forgiving of sins without bloodshedding, so fast get all things cleansed by the
blood’ (FgR p.101)

Once again, fast precedes a quantified NP. A verb-modifier reading yields the proposi-
tion ‘all things get very much cleaned by Christ’s blood’. However, the theological
notion of being clean from sins is not gradable, and the proposition therefore implausi-
ble. In terms of linguistic structure, fastnew yields a plausible reading: [fastnew + all things]
get cleaned by Christ’s blood. This message, however, would be even more heretical,
as no sin is too bad to be cleansed by Christ’s blood. The author hence must have
intended a proposition oriented reading: ‘as much as no sins can be forgiven without
Christ’s blood’, ßo fast = so much ⁄ certainly will all things be purified by the blood.’
The author attests that the proposition ‘all things get cleansed by Christ’s blood’ a high
degree of truth.
While this may be logically senseless, the concept exists in many languages. Interest-
ingly, the notion of something being ‘very true’ can apply to non-gradable as well as
gradable content as illustrated by an example from ModE:

(10) As much as the Queen is the leader of the UK, she still has to obey the law.
‘It is very true that the Queen is the leader, but still …’

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
42 Regine Eckardt

I will not expand this into a full analysis, but the construction offers a possibility to com-
bine ‘very much’ with sentences that do not themselves refer to grades. More quotes of
this kind are added in an appendix.
The quotes from FgR show that speakers ⁄ writers used fastold in sentences that were
quite varied (and hence don’t exhibit one uniform type of construction) but all share the
fact that they were somewhat unperspicuous. Specifically, it is propositional uses in the
sense of ‘it is very true that p’ which often contain fastold with a sister constituent that
doesn’t denote a gradable concept. The writers most likely intended, and readers could
still interpret the utterence based on fastold. Alternatively, at least some of these cases
could also lend themselves to a semantic composition with fastnew. A reader who saw this
alternative way of semantic composition could hence believe that other speakers used fast
in a new sense that was not yet part of his ⁄ her own mental lexicon. We simply need to
find examples that could suggest a new meaning fastnew = ‘almost’ to such uncertain hear-
ers ⁄ readers. Let me offer several suggestions about potential places for reanalysis.
In some cases, negation of high degrees results in an ‘almost’ reading. The following
quote exemplifies the constellation.

(11) es ist nit fast nutz …

it is not fast useful
(FgR, p. 558)

A conservative narrow scope LF [‘not’ [fast old + ‘useful’]] yields the statement that some-
thing is ‘not very useful’.
[[es ist nit fast nutz]]old = ‘its degree of usefulness is not very high’

The reader might understand the stronger message that the thing in question is ‘almost
useless’. This message can feed a reanalysis into the new LF [fastnew + [‘not useful’ ⁄
‘useless’]].

[[es ist nit fast nutz]]new = ‘it is almost useless’

reanalysing [[fast]]new = kP.ALMOST(P)

Another option could be reanalysis triggered by scalar implicatures. Intensifying fast is

often combined with other intensifying constructions, like in the following example:

(12) sei rouften laut, sei schrewend fast

they called loudly, they screamed fast
Heinrich Wittenwiler, Der Ring (c. 1408 ⁄ 10), part I, l. 352

This example is pragmatically precarious. Arguably, schrewen (‘scream’) is a more intense

variant of roufen (‘call’). roufen can be intensified to rouften laut which, by itself, gives rise
to the scalar implicature ‘they didn’t do anything more intense than call loudly’. The sub-
sequent statement that they also schrewend fast (‘scream very much’) can be interpreted
conservatively, and hence cancel the implicature. However, the reader ⁄ hearer could also
be led to assume an alternative rhetorical structure in which the second conjunct elabo-
rates the first:

(12¢) they cried loudly, they almost screamed

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
 Semantic Reanalysis and Language Change 43

If this were the understood message, the hearer would again infer a (new) lexical entry
[[fast]]new =kP.ALMOST(P).
Finally, there could be reanalysis from the propositional ‘as much as it is true that …’
operator to an ‘almost’ operator which only takes low scope over its argument all ‘all’.
The following quote illustrates this. In the present example, reanalysis exploits a discours
use of so = ‘given that’.

(13) wa sol ich dich aber uff daz kürtzst hinschicken in

where shall I you but on the shortest send in
Paulo, Jacobo, Petro, Johanne, so fast all ir episteln uff
Paulus, Jacob, Petrus, John, so fast all their epistles on
gesetze und guote werck gen?
laws and good deeds go
(FgR, p. 408)

The author wants to back up an earlier claim by a supporting quote from the New Tes-
tament which, as stated, contains plenty: which one should he pick? In a conservative
interpretation, the author wrote ‘where should I direct you – as much as it is true that all
epistles are about laws and good deeds?’ i.e. a rhetorical gesture of helplessness. (Lutherans
were commonly better read in the bible than Catholics.) An alternative interpretation
would essentially make the same point: ‘Where should I direct you, given that almost all
their epistles aim towards laws and good deeds?’.
LFold: [so fast [all epistles go for laws and good deeds]]

[[S]]old = ‘as much as it’s true that p’

p = ‘all their epistles aim towards laws and good deeds’

LFnew: [so [[fast all epistles] go for laws and good deeds]]

[[S]]new =‘as (almost all) their epistles aim towards laws and good deeds’
reanalysing [[fast]]new = kP.ALMOST(P)

Potential sources for reanalysis might have occasionally occurred at all times, but they
gained momentum when the general spectrum of uses of fast in all its older senses had
become wide, and when precarious (older) uses seem to increase.9 I took this final,
lengthy example to illustrate not only semantic reanalysis as a process, but also the linguis-
tic and communicative landscape in which semantic reanalysis occured.

4. More Research, Past and Future

This article intended to show how hearers arrange and re-arrange meanings when they
interpret an utterance. Such re-arrangements can lead to persisting language change.
Semantic reanalysis is an independent mode of semantic change, next to other, better-
described changes like metaphoric extension, metonymy, narrowing or broadening.
Semantic reanalysis is one more phenomenon in languages which wittnesses speakers’ trust
in semantic composition. With its focus on composition, the approach allows us to better
understand the unconscious reasoning of the hearer ⁄ reader in the phase of change, as well
as the communicative situations that spur such reasoning. I suggested that precarious uses

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
44 Regine Eckardt

of words can invite hearers to reanalysis. More cases illustrate this point, like lauter (Eckardt
2006), even (Traugott 2006, Eckardt 2009), selbst (Eckardt 2006) and it is indirectly advo-
cated by authors who warn against teleological models of change: language change is not
an automatism which carries languages towards an ideal end state.
Semantic and pragmatic formats that reflect semantic composition at an appropriate
level of detail are indispensible in understanding the laws and limits of semantic reanalysis.
Formal analyses can help to separate instances of generalization from instances of reanaly-
sis (see Condoravdi and Deo 2008), investigate the semantics of functional terms in detail
(Deo 2006; Gerö and von Stechow 2003; Haug 2008) and, finally, begin to understand
the semantic and pragmatic contribution of grammaticised discourse markers (Zeevat and
Karagjosova, 2009).
Such analyses also newly pose the question about pushing factors in language change.
Researchers have suggested many reasons that might drive speakers ⁄ hearers to reform
their internal grammars, ranging from laziness, or the need to increase the expressive
power of one’s language, over the desire to be original to the urge to simplify and sys-
tematize a local part of one’s grammatical system. The present article is modestly
restricted to the discussion of one particular mode of semantic change and I will therefore
not attempt to weigh the reasons that spur semantic reanalysis against other driving forces
of change. Taking a wider perspective, note that precarious language use can be precari-
ous in many ways. It can challenge the hearer by pragmatic infelicities, by too rich
background assumptions, by uncommon or archaic syntactic constructions, unusual opera-
tor-argument constellations in syntax, semantically baroqueness and maybe more.10 What
remains typical, however, is that the semantic changes that affect parts of the utterance
are determined ‘top-down’ by the hearer’s attempt to understand how the overall message
is built up in the sentence. These changes are therefore not necessarily driven by the
semantic qualities inherent to the word or expression, e.g. by coercion or metpahoric
potential. The changes can be minor, but can also span surprising conceptual gaps: For
instance, Botne (1998) reports on Bantu languages where the verb say developed a second
use as a future tense serial verb (something similar to an auxiliary, but in more complex
interaction with the main verb of the sentence). It will be desirable, hence, to accumulate
a representative range of detailed case studies before we can make out universal trends
like semantic simplification, streamlining of the mapping between syntax and semantics,
improvement of syntactic elegance or restructuring in analogy to other dominant patterns
in the language, and before we can say which ones predominate as causes in change.

Short Biography
Regine Eckardt holds a doctorate from Stuttgart and is now Professor of Linguistics in
the Department of English Language and Literature at Göttingen University. Building on
her early research topics in synchronic semantics and pragmatics, she has been working
on meaning change and grammaticalization since 2000. Specifically, she pursues the
hypothesis that the paradigm of truth conditional, compositional semantics and pragmatics
is particularly suited to analyse the shifts and changes of semantic composition in gram-
maticalization.

Notes
* Correspondence address: Regine Eckardt, English Department ⁄ Linguistics, University of Göttingen, Käte-
Hamburger-Weg 3, D-37073 Göttingen, Germany. E-mail: [email protected]

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
 Semantic Reanalysis and Language Change 45
1
The field will be comprehensively covered in Heine, B. and H. Narrog (eds.), Handbook of Grammaticalization,
to appear at Oxford University press.
2
To be precise, I should mention that the positions on clines can often also include a syntactic characterization of
the respective word(s). Being mostly concerned with meaning change here, I left these out. On the other hand, each
semantic positions on a cline is usually loosely taken and might cover more than one denotation in, e.g. the strict sense
of truth conditional semantics. Otherwise, chances are dim that any cline could actually be travelled more than once.
3
Resort to non-compositional interpretation can happen, but for the sake of the discussion, I will stick to the
conservative assumption that it doesn’t in the cases at hand.
4
Readers of Diewald (2002) will acknowledge that precarious contexts don’t coincide with her atypical contexts
either because these are characterized by standard pragmatic implicatures, not by pragmatic overload or markedness.
Diewald (2006) is written against a background of construction grammar, and assumes that word and sentence
meanings are more often non-compositional than not. I will not follow this assumption but have to leave a full jus-
tification for semantic compositionality to another occasion.
5
formally, p = kw.$e (CONFERENCEW(we,e) ! s(e) = 3h).
6
formally, p = kw.$e (CONFERENCEW(we,e) ! s(e) = ALMOST(3h)).
7
‘The author conveyed exactly ⁄ almost at two o’clock because here, exactness borders easily to vagueness’.
8
An anonymous reviewer suggests that complex clause structure could play a role in this change. I currently see no
direct connection, except perhaps the obvious correlation between overly complicated contents and overly compli-
cated syntax. Any such hypothesis, hence, will have to be subject to future research.
9
It might be tempting to speculate that the speaker-hearer constellation in this particular type of text, written by
theological experts for lay readers, likewise supported reanalysis. A verification of sociolinguistic claims requires
electronic sources that are not, presently, available for German.
10
See Roberts and Roussou 2003 and Roberts 2010 for a recent discussion of cases where morphosyntactic facili-
tation is hypothesized to drive language change.

Works Cited
Amaral, Patricia Matos. 2006. The polysemy of ‘mal’ in European Portuguese: a diachronic analysis. Journal of
Historical Pragmatics 7. 1–37.
Behaghel, Otto. 1924. Deutsche Syntax. Vol. II. Heidelberg: Winter.
Botne, Robert. 1998. The evolution of future tenses from serial ‘say’ constructions in central eastern Bantu. Diach-
ronica XV.2. 207–30.
Condoravdi, Cleo, and Ashwini Deo. 2008. Aspect shifts in Indo-Aryan. Proceedings of the 18th International Congress of
Linguists (CIL 18): workshop on formal approaches to the relation of tense, aspect and modality, 21–26 July, Seoul.
Deo, Ashwini. 2006. Tense and aspect in Indo-Aryan languages: variation and diachrony. Yale University, New
Haven: PhD dissertation.
Diewald, Gabriele. 2002. A model for relevant types of contexts in grammaticalization. New reflections on gram-
maticalization, ed. by Ilse Wischer and Gabriele Diewald, 103–20. Amsterdam: John Benjamins.
——. 2006. Context types in grammaticalization as constructions. 15th Sept. 2010 <http://www.constructions-onli-
ne.de/articles/specvol1/686>.
Eckardt, Regine. 2006. Meaning change in grammaticalization. Oxford: Oxford University Press.
——. 2007. Almost — an analysis. 17th Sept. 2010 <http://www.semanticsarchive.net/>.
——. 2009. APO—avoid pragmatic overload. Current trends in diachronic semantics and pragmatics, ed. by
Maj-Britt Mosegaard Hansen and Jacqueline Visconti, 21–42. UK: Emerald.
——, Klaus v. Heusinger, and Christoph Schwarze (eds). 2003. Words in time. Berlin: Mouton deGryuter.
Fischer, Olga. 2007. Morphosyntactic change: functional and formal perspectives. Oxford: Oxford University Press.
Gelderen, Elly van. 2008. Economy of merge and grammaticalization. Variation, selection, development, ed. by
Regine Eckardt, Gerhard Jäger and Tonjes Venstra, 179–98. Berlin: Mouton deGruyter.
Gerö, Eva-Catrin, and Arnim von Stechow. 2003. Tense in time: the greek perfect. Words in time, ed. by Regine
Eckardt, Klaus v. Heusinger and Christoph Schwarze, 251–94. Berlin: Mouton deGryuter.
Grimm, Jakob, and Wilhelm Grimm (eds). 1854–1960. Deutsches Wörterbuch. Leipzig: S. Hirzel.
Haug, Dag. 2008. From resultatives to anteriors in Ancient Greek. On the role of paradigmaticality in semantic change.
Grammatical change and linguistic theory, ed. by Th. Eythorsson, 285–305. Amsterdam: John Benjamins.
Heine, Bernd. 2002. On the role of context in grammaticalization. New reflections on grammaticalization, ed. by
Ilse Wischer and Gabriele Diewald, 83–102. Amsterdam: John Benjamins.
——, and Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press.
Hopper, Paul, and Elizabeth C. Traugott. 2003. Grammaticalization. Cambridge: Cambridge University Press.
Second, revised edition.
Kennedy, Chris, and Louise McNally. 2005. Scale structure, degree modification, and the semantics of gradable
adjectives. Language 82(2). 345–81.

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd
46 Regine Eckardt

Kytö, Merja, and Suzanne Romaine. 2005. We had like to have been killed by thunder & lightning: the semantic
and pragmatic history of a construction that like to disappeared. Journal of Historical Pragmatics 6. 1–35.
Lightfoot, David. 1991. How to set parameters: arguments from language change. Cambridge, MA: MIT Press.
Penka, Doris. 2006. Almost there: the meaning of almost. Proceedings of Sinn und Bedeutung 10. ZAS papers in
linguistics, Nr. 44, ed. by C. Ebert and C. Endriss, 275–86. Berlin: ZAS.
Roberts, Ian. 2010. Grammaticalization, the clausal hierarchy, and semantic bleaching. Gradience, gradualness, and
grammaticalization, ed. by Elizabeth C. Traugott and Graeme Trousdale, 45–73. Amsterdam: Benjamins.
——, and Anna Roussou. 2003. Syntactic change: a minimalist approach to grammaticalization. Cambridge: Cam-
bridge University Press.
Traugott, Elizabeth C. 1988. Pragmatic strengthening and grammaticalization. Proceedings of the 14th annual meet-
ing of the Berkeley Linguistic Society, ed. by S. Axmaker, A. Jaisser and H. Singmaster, 406–16. Berkeley:
Berkeley Linguistic Society.
——. 2006. The semantic development of scalar focus particles. The handbook of the history of English, ed. by
Ans van Kemenade and Bettelou Los, 335–59. London: Blackwell.
——, and Richard B. Dasher. 2002. Regularity in semantic change. Cambridge: Cambridge University Press.
Zeevat, Henk, , and E. Karagjosova. 2009. The history and grammaticalization of ‘‘doch’’ ⁄ ’’toch’’. ZAS papers in
linguistics Nr. 51, Berlin: ZAS.

Source Texts
Das Bonner Frühneuhochdeutsch-Korpus. 18th Aug. 2010 <http://www.korpora.org/fnhd/>.
Laube, A., and U. Weiss (eds). 1997. Flugschriften gegen die Reformation (1518 – 1524). Berlin: Akademie Verlag.
20th Aug. 2010 <http://books.google.de>.
Wittenwiler, Heinrich. (c. 1408/10). Der Ring, Part I ‘‘Die Brautwerbung’’. Quoted after Bibliotheca Augustana.
20th Aug. 2010 <http://www.hs-augsburg.de/~harsch/germanica>.

Search Covered
Altdeutsche Predigten I, Obersächsisch, Early 14. century
Benediktinerregel Oxford, Nassau 14. century
Buch Altväter, Stuttgart 14. Jahrhundert
Das Buch der Cirurgia des Hieronymus Brunschwig. Straßburg, 1497
Des Gottesfreundes im Oberland (= Rulmann Merswin’s) Buch von den zwei Mannen. Nach der ältesten
Strassburger Handschrift 1352
Gerold Edlibach: Chronik, Zürich 1485–1486
Gualtherus Burlaeus: Das Buoch von dem Leben und Sitten der heydnischen Maister, Augsburg 1490
Handschrift Pillenreuth Mystik, Nürnberg 1463
Hans Mair: Troja, Nördlingen 1392
Hans Neidhart: Eunuchus des Terenz, Ulm 1486
Helene Kottanerin: Denkwürdigkeiten, Wien 1445–1452
Johannes Rothe: Chronik, Thüringisch second half, 15. century
Johannes Tauler: Sermon des grosz gelarten, in Gnade erlauchte Doctoris Johannis Thauleri Predigerr Ordens,
Leipzig 1498
Johann Wonnecke von Cube: Hortus Sanitatis, Mainz 1485
Mönch von Heilsbronn: Von den sechs Namen des Fronleichnams, Nürnberg late 14. century
Naturlehre Mainau, Osthochalemannisch, late 14. century
Wilhelm Durandus: Rationale, Wien 1384
Zwei Psalter aus dem 14. Jahrhundert (Dresden MS. 287 und Hamburg in SCR. 142) und drei verwandte
Bruchstücke aus Schleiz, Breslau und Düsseldorf

ª 2011 The Author Language and Linguistics Compass 5/1 (2011): 33–46, 10.1111/j.1749-818x.2010.00260.x
Language and Linguistics Compass ª 2011 Blackwell Publishing Ltd