![Type entrenchment vs. productive instantiations of [fy fa-x]](https://www.researchgate.net/profile/Steffen-Hoeder/publication/374782920/figure/fig1/AS:11431281199472489@1697615375888/Type-entrenchment-vs-productive-instantiations-of-fy-fa-x_Q320.jpg)
![Lexical association of fan and helvete with [fy sw]](https://www.researchgate.net/profile/Steffen-Hoeder/publication/374782920/figure/tbl2/AS:11431281199472491@1697615376112/Lexical-association-of-fan-and-helvete-with-fy-sw_Q320.jpg)
![Lexical association of fan and helvete with [sw också]](https://www.researchgate.net/profile/Steffen-Hoeder/publication/374782920/figure/tbl3/AS:11431281199472492@1697615376274/Lexical-association-of-fan-and-helvete-with-sw-ocksa_Q320.jpg)
Content uploaded by Steffen Höder
Author content
All content in this area was uploaded by Steffen Höder on Oct 18, 2023
Content may be subject to copyright.
The Devil is in the schema
A constructional perspective
on Swedish taboo-avoiding strategies
Steen Höder
Christian-Albrechts-Universität zu Kiel
Swedish swearwords are predominantly religious in origin (e.g. fan ‘the
Devil’, helvete ‘hell’, and jävlar ‘devils, demons’). The former taboo status of
swearing is still reected in the existence and productive use of taboo-
avoiding strategies, most notably phonological modication (e.g. fasen <fan
‘the Devil’, helsicke <helvete ‘hell’). This paper discusses such taboo-
avoiding strategies from the perspective of usage-based Construction Gram-
mar. It argues that taboo-avoiding relies on schematic swearing
constructions in combination with radical coercion in general and on sub-
morphemic coercion in particular: The meaning of a taboo-avoiding
expression is entirely constructional, and the lexical semantics of the slot-
lling items is irrelevant. Evidence for the cognitive reality of phonologi-
cally schematic swearing constructions is found in a corpus-based analysis
of selected taboo-avoiding patterns, which shows that swearing construc-
tions are not only instantiated by lexicalized variants, but also used produc-
tively, as illustrated by infrequent types.
Keywords: swearing, linguistic taboo, Swedish, Construction Grammar,
phonological schematicity
. Introduction
While swearing as a communicative practice seems to be universal, conventions
dier across speech communities as to which types of linguistic elements are con-
sidered swear words (SWs), to what degree swearing is considered oensive, and
how and when speakers avoid the kind of linguistic taboo (cf. Allan & Burridge,
) that SWs might carry. For example, while both the use of English fuck and
Swedish knull ‘fuck’ is considered vulgar by most speakers, only fuck is used as
a common SW, whereas knull is exclusively used referentially. Similarly, German
https://doi.org/10.1075/cal.37.04hod
© 2024 John Benjamins Publishing Company
Sakrament ‘sacrament’ is used as a SW in Bavarian, but not in other regional vari-
eties of German. Finally, while Swedish jävlar ‘devils, demons’ is one of the lan-
guage’s most common SWs, it is not so oensive as to be considered unacceptable
on television or radio, in contrast to, for example, the United States, where it is
common practice to bleep out similarly frequent SWs, or to asterisk them out in
writing (cf. what the f**k is she doing?).
The latter example represents a taboo-avoiding strategy (TAS) that is fairly
common: the substitution of oensive lexical material with non-verbal sounds
(in spoken language) or symbols (in written language). In spontaneous speech,
speakers oen employ similar TASs that are based on the substitution of oensive
words with dierent lexical material, sometimes referred to as euphemistic substi-
tution (Casas Gómez, ). For example, speakers of German sometimes replace
Teufel ‘Devil’ with Kuckuck ‘cuckoo’ in expressions such as was zum Kuckuck
(instead of was zum Teufel ‘what the hell’). Oen, the euphemistic variants are
phonologically similar to the original taboo words, as in the English forms darn,
shoot, and heck used instead of damn,shit, and hell. In many languages, these TASs
are conventionally xed, with a limited set of lexical elements that are used as
stand-ins for others that are (more) taboo. In other languages such as Swedish,
TASs are used far more productively, resulting in a plethora of euphemistic sub-
stitutes, ranging from well-established and frequent forms like sjutton ‘seventeen’
or farao ‘pharaoh’ (both replacing fan ‘the Devil’) to more ad hoc formations such
as helsinki ‘Helsinki [Finland’s capital, normally called Helsingfors in Swedish]’
(replacing helvete ‘hell’), as in ().
() många
many
av
of
er
you.
undrade
wondered
vad
what
i
in
helsinki
helsinki
som
.
hade
had
hänt
happened
‘many of you wondered what the hell [literally: what in Helsinki] had hap-
[Bloggmix : My Baby Dolls, ––]pened’
What makes oensive words taboo is, rst and foremost, their form, not their
function – otherwise, euphemistic substitution would not be of any use. This is
also what makes TASs an interesting phenomenon from a constructionist per-
spective: If the genuine, conventional form of a SW is replaced with a non-
conventional variant, then what is it that conveys its meaning, and how does it do
that? More specically, how can the productivity of TASs be modelled in a way
that is cognitively plausible and ties in with current models of linguistic knowl-
edge and usage, in particular with respect to central notions such as cognitive
entrenchment, as measurable by means of corpus data?
This chapter tackles these questions from a usage-based Construction Gram-
mar (CxG) perspective. Its aim is twofold:
Steen Höder
First, it proposes a constructionist model of TASs in terms of schematic con-
structions: It is argued that TASs rely on the existence of schematic swearing
constructions that, in combination with contextual cues, provide enough infor-
mation for euphemistic substitutions to be decoded as intended. Furthermore, it
is argued that an additional type of constructional cue is provided by submor-
phemic patterns that can be described in terms of phonological schemas, similar to
schematic sound patterns that are assumed for constructionist models of, among
other things, non-concatenative morphology (cf. Höder, , pp.–). Both
kinds of schemas work through dierent, but related types of coercion, one that
has been described as radical coercion (resulting in exclusively construction-
determined meaning; Audring & Booij, ) and one that involves phonological
creativity on a submorphemic level (submorphemic coercion).
Second, an analysis of present-day Swedish corpus data is presented which
provides evidence that phonological schemas of the proposed type are cognitively
realistic and are used productively in informal speech, albeit to varying degrees
(in line with the use of frequency and productivity measures as evidence in usage-
based linguistics; Bybee, , p.). Furthermore, the results suggest that such
phonological schemas are utilized, at least by some speakers, as a resource for lin-
guistic creativity and extravagance (for a discussion of these concepts, cf. Bergs &
Kompa, , Haspelmath, , and Ungerer & Hartmann, ; what is under-
stood as creativity here falls into the realm of xed creativity in Berg & Kompa’s,
, terminology).1
The data are taken from the Bloggmix and Diskussionsforum corpora, in total
corpora with . billion tokens (as of ––; Bloggmix: corpora with
approximately . billion tokens, Diskussionsforum: corpora with approxi-
mately . billion tokens).2These corpora, which consist of selected posts from
Swedish blogs and online forums, represent informal text types that occupy an
intermediate position between informal spoken language and more prototypically
written registers. Hence, they are particularly useful when studying phenomena
. In Bergs & Kompa’s, , terminology, ‘F-creativity’ (‘xed creativity’) is understood as
making maximum use of linguistic possibilities rather than breaking out of structural patterns
altogether. Their recent interpretation of this term is dierent from earlier constructional
approaches to creativity (Bergs, ), which would instead place (some) TASs into the category
of ‘E-creativity’ (‘extending creativity’).
. Both corpora are accessible through a common interface, Korp (spraakbanken.gu.se/korp),
which is maintained by Språkbanken Text, a Swedish research and language technology unit
located at the University of Gothenburg, part of Nationella Språkbanken (The National Lan-
guage Bank) (Borin, Forsberg & Roxendal, ). Corpus queries can be formulated in CQP
(Corpus Query Processor), an advanced query language that, among other things, supports the
use of regular expressions (Evert ).
Chapter . The Devil is in the schema
that are rather typical of informal speech. Of course, it would be desirable to (in
addition) use actual spoken language data, but since – as of now – there are no
corpora of spoken Swedish large enough so as to be useful for analyses of infre-
quent phenomena, this study had to be restricted to written data.
The chapter is structured as follows: Section delves into the forms and func-
tions of swearing in present-day Swedish as well as its cultural history. This sets
the scene for the discussion in the subsequent sections, which tackle Swedish
TASs from a constructional perspective. Section discusses euphemistic substi-
tution and the constructional mechanism of radical coercion, including cases of
submorphemic creativity, while Section provides a quantitative analysis of cor-
pus data that sheds light on the productivity of related constructions, before con-
clusions are reached in Section .
. Swearing in Swedish: Cultural history, forms, and functions
Swearing is, as far as we can tell, used across all languages and speech communi-
ties (for an extensive survey of the history of swearing, cf. Ljung, , pp.–),
and it is by no means a marginal phenomenon, but – like other types of taboo lan-
guage (cf. Culpeper, ) – occurs rather frequently, primarily in informal, col-
loquial speech (Jay, ; Beers Fägersten & Stapleton, , pp.–). Based on
earlier research, Ljung (, pp.–) provides and discusses a useful set of crite-
ria for a working denition of swearing. These include the following (a–c).
a. Utilization of taboo words: Swearing involves lexemes that, conventionally,
qualify as inappropriate in a given communicative setting, either because they
are considered inappropriate in general (as with fuck) or because their appro-
priate use is conventionally restricted to specic domains (as with hell, which
is acceptable in, say, theological contexts).
b. Expressive function: Swearing expresses the speaker’s feelings in a specic
communicative setting or their attitudes towards one or more referents of an
utterance. This covers negative emotions such as pain or fear as well as neu-
tral or positive ones such as surprise and joy.
c. Non-referential meaning: While SWs are homonyms of lexemes that carry ref-
erential meaning, they are not used referentially; this is what sets them apart
from other types of oensive words (e.g. shit! used as an emotive exclamation
qualies as a SW, whereas shit as a vulgar expression referring to faeces does
not).
Steen Höder
As a working denition, swearing can thus be characterized as the expressive use
of conventionally taboo lexical elements without referential meaning.3
Speech communities draw on dierent source domains for swearing, at least
in part depending on dierent traditional values in individual societies (cf. Ljung,
, pp.–). Nübling & Vogel’s (; cf. Table ) comparison of typical SWs
in Dutch, German, and Swedish demonstrates that, for example, Dutch makes
use of sexual, scatological, religious, and disease-related themes, while the disease
theme is virtually absent from German and Swedish. Also, dierent languages
exhibit dierent preferences for existing themes: While sexual and, to a lesser
extent, disease-related SWs are dominant in Dutch, the scatological theme is
prevalent in German. Most crucially for this paper, Swedish SWs are predomi-
nantly religious in origin (Nübling & Vogel, ; the same holds true, at least
traditionally, for other Scandinavian languages; cf. Fjeld, ; Rathje, ; for a
historical perspective, cf. Stroh-Wollin, , ), with forms of diabolic words
such as fan ‘the Devil’, satan ‘Satan’, helvete ‘hell’, djävul or jävel ‘devil, demon’
(oen in the plural jävlar or the historic genitive plural jävla) among the most fre-
quent ones, while scatological SWs such as skit ‘shit’ and piss ‘piss’ are used less
oen; sexual SWs appear to be on the rise, though, mainly due to English loan-
words such as fuck.
Table . Source domains of SWs in Dutch, German, and Swedish
(Nübling & Vogel, , p. ; own translation)
sexual scatological religious disease-related
Dutch +++ +(+) +(+) ++(+)
German [+] +++ ++ –
Swedish [+] + +++ –
It has sometimes been argued that language-specic theme preferences mir-
ror culture-specic taboo hierarchies, i.e. preferred themes correspond to source
domains that are (or have been) strongly taboo. Even though such a strong claim
does not appear to be well substantiated (Nübling & Vogel, , pp.–), it is
obvious that the cultural background does play some role for the perception of
. The third criterion is slightly problematic because, etymologically, most SWs are (or can be
traced back to) conventionalized metaphors (e.g. damn[ed] ‘as if [literally] damned’) or seman-
tically bleached expressions (e.g. hell ‘some bad place’), and there are ambiguous contexts (e.g.
it burned like hell ‘it burned very much [non-literal]’ or ‘it burned like hell burns [literal]; cf.
Ljung, , p.; Stroh-Wollin, , p.). In general, however, SWs are typically fully con-
ventionalized and do not carry referential meaning (any more).
Chapter . The Devil is in the schema
linguistic taboo and the importance of potential source domains for the conven-
tionalization of SWs. The dominance of religious SWs in Swedish has its roots in
the Christian cultural heritage of the Swedish-speaking countries (cf. Andersson,
, pp.–; Hedin, ).
Traditionally, Sweden (until including present-day Finland) was a
monoconfessional Lutheran society, with the Swedish Church as its established
national church. Conservative protestant thinking remained inuential until at
least the th century, which saw an increase in religious freedom (with the
repeal of the Conventicle Act [konventikelplakatet] that had made private
religious gatherings illegal and the / enactment of the Dissenter Acts
[dissenterlagarna] that under specic conditions legalized leaving the national
church).4The church also exerted intellectual control over the population
through the yearly examination of residents’ competences (in particular biblical
knowledge, familiarity with the Catechism, and reading skills) by the parish
priest (husförhör, abolished ). In addition to mainstream Christianity, non-
conformist – mainly pietistic and revivalist – movements gained inuence from
the th century onwards. All of this contributed to the entrenchment of religious
thought, education, ritual, and piety in the social practices and norms of Swedish
society. Moreover, both traditional Christianity and folk belief entailed a certain
degree of magical thinking. For example, hell and the Devil (as well as other
supernatural beings that belong to Nordic folklore) were, by many if not most,
thought of as real entities that existed in the actual world and that people could
interact with in various ways. This included the possibility of summoning them or
invoking their actions by calling them by their names, as embodied in the Swedish
proverb när man talar om trollen [så står de i farstun] ‘when you talk about the
trolls [they are standing at your doorstep]’, equivalent to English speak of the
Devil [and he shall appear] (cf. Andersson, , p.). At the same time, actu-
ally doing this was considered taboo if not downright blasphemous, in particu-
lar when appealing to God under inappropriate circumstances; this was explicitly
forbidden in the Second of the Ten Commandments (according to the numbering
in Luther’s Small Catechism): du skall icke missbruka Herrens, din Guds, namn
‘thou shalt not take the name of the Lord thy God in vain’.
. The historically strong role of the church can be further illustrated by the fact that the
unconditional right to leave the national church was not granted until the Religious Freedom
Act (religionsfrihetslagen) of . Independently of individual religious aliation, the Swedish
Church remained responsible for resident registration in parish registers until , and the o-
cial separation of church and state was not completed until , when the Swedish Church
was nally disestablished.
Steen Höder
In today’s secularized society, swearing is not viewed as religiously oensive
any longer. Religiously based reservations against swearing were virtually inexis-
tent as early as in the s (Stroh-Wollin, , pp.–), although SWs may still
be considered inappropriate depending on speakers’ attitudes as well as commu-
nicative settings. Unsurprisingly, however, religious swearing is still fairly com-
mon in colloquial present-day Swedish. This may be illustrated by the frequencies
of the most common SWs in the Bloggmix and Diskussionsforum corpora: There
are about tokens of fan per million words ( tpm), tpm of helvete, tpm
of jävlar. Even if swearing can hardly be considered taboo in the traditional sense
any more, its former taboo status is still reected in the existence and frequent use
of a range of TASs, which are employed as a means of both euphemistic substi-
tution and linguistic creativity, even if TASs are less frequent than actual swear-
ing. Frequent TASs include the lexical substitution of diabolic words such as fan
‘the Devil’ with common nouns (e.g. katten ‘the cat’) and numerals or numeral-
derived elements (e.g. sjutton ‘seventeen’, attan(s) <arton ‘eighteen’, hundan <
hundra ‘hundred,’ tusan <tusen ‘thousand’) as well as, most notably, phonologi-
cal modication, which sometimes entails the use of pre-existing, phonologically
similar words (e.g. farao ‘pharaoh’), but oen also results in non-canonical forms
that are not homonyms of pre-existing lexemes (e.g. fasiken; cf. Teleman, ,
pp.–; Stroh-Wollin, , pp. –).
. A constructional approach to hell: The devil is in the schema
. Swearwords and swearing constructions
While SWs can be used in the same way as other lexemes (of the same word
classes), there is also a cross-linguistic tendency for SWs to occur within formulaic
sequences, i.e. in specialized constructional contexts that can be rather idio-
syncratic (e.g. who/what/where the devil/fuck; Ljung, , p.). Consequently,
Andersson (, pp.–), Teleman () and Hoeksema () explicitly
speak of ‘the grammar’ of SWs (and other taboo elements).
More specically, although most Swedish SWs are etymologically nouns, they
occur in a surprisingly wide range of non-nominal contexts as well. From a tra-
ditional perspective, Swedish SWs can be said to occur in dierent syntactic
positions and grammatical functions, as illustrated in Table (cf. Teleman, ,
pp.–), including extraclausal uses as in () and (), dierent types of intensi-
ers as in () and (), emphatic question particles as in (), and suxoids (bound
morphemes that can be analysed either as head elements within a compound or
as derivational suxes) as in ().
Chapter . The Devil is in the schema
Table . Grammatical functions of Swedish SWs
Class Examples
() interjections a. Fan!
Devil.the
‘Fuck!’
b. Jävlar!
devils
‘Damn!’
() interjectional phrases a. Fy
phew
fan!
Devil.the
‘Fuck!’
b. Jävlar
devils
också!
also
‘Damn!’
c. För
for
helvete!
hell
‘Hell!’
() adjectival intensiers a. din
your
jävla
devils’
idiot
fool
‘you bloody fool’
b. ett
one
satans
Satan’s
mål
goal
‘one bloody goal’
() adverbial intensiers a. Det
it
är
is
jävla
devils’
kallt.
cold
‘It’s fucking cold.’
b. Det
it
var
was
billigt
cheap
som
as
fan.
Devil.the
‘It was fucking cheap.’
c. Inte
not
fan
Devil.the
är
am
jag
I
nazist.
Nazi
‘I’m not a fucking Nazi.’
() emphatic question particles a. Vad
what
fan
Devil.the
håller
hold
du
you
på
on
med?
with
‘What the fuck are you doing?’
b. Var
where
i
in
helvete
hell
har
have
du
you
varit?
been
‘Where the hell have you been?’
c. När
when
jävlar
devils
ska
shall
man
one
hinna?
have.time?
‘When the fuck are you supposed to have time?’
Steen Höder
Table . (continued)
Class Examples
() suxoids a. bilhelvetet
car-hell-the
‘the bloody car’
b. kattjäveln
cat-devil-the
‘the fucking cat’
From a constructionist view, this corresponds to a network of at least partially
lled constructions (cxns) that are interconnected by both horizontal and vertical
links (to adopt the perspective advocated by, amongst others, Diessel, ). Inter-
jections consisting of a single SW can be modelled as individual cxns (such as
[fan] or [jävlar]) in addition to a schematic swearing cxn (SCxn) that denes
them as a class ([.]), i.e. captures their distributional similarities
such as their ability to form a sentence (with or without an initial conjunction,
as in och/eller/men fan ‘and/or/but fuck [lit.: and/or/but the Devil]’). Another
relevant schema is [, .], which is instantiated by SCxns
with preposed particles such as [fy ] (as in fy fan ‘fuck [lit.: phew the Devil]’)
or [för ] (as in för helvete ‘hell [lit.: for hell]’) or postposed particles such as
[ också] (as in jävlar också ‘damn [lit.: devils also]’). The use of SWs as inten-
siers corresponds to partially schematic SCxns such as [jävla ] (e.g. jävla
idiot ‘bloody fool [lit.: devils’ fool]’), [jävla ] (e.g. jävla kallt ‘fucking cold [lit.:
devils’ cold]’), [ som fan] (e.g. billigt som fan ‘fucking cheap [lit.: cheap as the
Devil]’), and [- fan] (e.g. inte fan ‘fucking not [lit.: not the
Devil]’). Some of those in turn instantiate fully schematic SCxns that contain SW
slots, such as [ ], which is instantiated by both [jävla ] and [satans
] (e.g. satans mål ‘bloody goal [lit.: Satan’s goal]’). Similarly, SWs used as
emphatic question particles can be modelled in terms of individual constructions
as well as a more general schema [. ] (e.g. vad fan ‘what the fuck [lit.:
what the Devil]’, när jävlar ‘when the fuck [lit.: when devils]’). Finally, SW suf-
xoids correspond to schemas such as [-helvete] (e.g. bilhelvete ‘bloody car
[lit.: car hell]’), [-jävel] (e.g. kattjävel ‘fucking cat [lit.: cat devil]’), which in
turn instantiate the schematic SCxn [-].
. Taboo-avoiding strategies, radical coercion, and extravagance
While the former taboo status of SWs was originally motivated by semantic prop-
erties of the individual lexical elements, such as their association with the diabolic
theme, this does not necessarily imply that there ever was a taboo about the
Chapter . The Devil is in the schema
semantics. This is indeed crucial: What had to be avoided was not referring to
the Devil as such, but referring to the Devil by a specic word or even a proper
name such as fan or satan. In constructional terms, then, SWs like fan originally
carry a referential meaning but are on top of that pragmatically marked as taboo;
in shorthand formalization: [fan ‘the Devil’ 〈taboo〉].5A cxn meant to avoid the
taboo therefore has to avoid specic forms as well as their pragmatic associations
while still encoding the intended meaning. However, it would be misleading to
assume that this leads to the establishment of some sort of lexical paradigm that
merely distinguishes tabooed and de-tabooed synonyms, as in [fan ‘the Devil’
〈taboo〉] vs. *[katten ‘the Devil’]. Firstly, lexical elements such as katten (lit. ‘the
cat’) are never used referentially to actually refer to the Devil, but only ever
expressively, i.e. as a SW. Secondly, this use of lexical elements as taboo-avoiding
equivalents presupposes some kind of constructional cue, in addition to the kind
of situational or contextual cues that always guide speakers’ interpretation of each
other’s utterances. In other words, the lexical element has to occur within a par-
ticular set of constructional schemas. For example, katten will not be decodable
as a taboo-avoiding variant of fan unless it lls a slot in, say, the [fy ] or the [
också] cxn (both instantiations of the more schematic [, .-
]), as in fy katten (lit.: ‘phew the cat’) or katten också (lit.: ‘the cat also’). This
implies that, say, the meaning of fy katten is not compositional, but entirely con-
structional: Whatever element lls the SW slot in [fy ] will be interpreted (and
in many cases accepted) as a euphemistic substitution. To put it dierently: The
Devil is in the schema, not in one of the words.
This mechanism represents a specic type of coercion. In CxG, this term in its
broadest sense refers to cases where the meaning of a slot-lling item (typically a
word) is, in some way, adjusted to the meaning of the larger constructional unit
(Michaelis, , p.). Coercion is particularly visible when the semantics of an
element is overridden – i.e. fundamentally altered, enriched, or suppressed – by
the meaning of the cxns in which it is embedded. Examples include the Individ-
uation Cxn (Hilpert, a, p.; Michaelis, , p.) as in () or the Caused-
Motion Cxn (Goldberg, , p. ) as in ().
() Four coees please!
. I adopt a maximalist view of meaning here that includes (conventional) pragmatics. This
is in line with general assumptions in the usage-based constructionist literature such as
Goldberg’s (, p.) denition of cxns as “emergent clusters of lossy memory traces that are
aligned within our […] conceptional space on the basis of shared form, function, and contextual
dimensions” [emphasis mine], but also ties in with more specic arguments in favour of includ-
ing pragmatic associations (e.g. Cappelle, ; Höder, ) and pragmatic or communicative
frames (e.g. Fischer, ; ; Czulo, Ziem & Torrent, ) on the functional side of cxns.
Steen Höder
() She sneezed the foam o the cappuccino.
In (), a mass noun (coee) is used as a count noun, given the right constructional
context provided by the numeral. In (), the interpretation as ‘She caused the
foam to move o the cappuccino by means of sneezing’ is derived from the con-
struction rather than one of the slot-lling elements.
In these prototypical cases, however, the lexical items nevertheless contribute
to the meaning of the whole construct. Thus, the Individuation Cxn has to be
combined with a (mass) noun specifying a substance (‘coee’); likewise, the ele-
ment that lls the verbal slot in the Caused-Motion Cxn species the manner of
causation (‘by means of sneezing’). Audring & Booij (, pp. –) argue
that this prototypical type of coercion (“coercion by override”) is taken to its logi-
cal extreme in what they label radical coercion: the mechanism by which a slot in a
schematic cxn is lled with a lexical element whose semantics contributes nothing
to the meaning of the utterance; its meaning is completely overridden by the cxn’s
meaning. Radical coercion is found in, for example, the German cxn [GEHEN,
nom, dat,auf acc ‘ annoys ’] as in (), where Zeiger ‘clock hand,
pointer’ and Sack ‘sack’ can ll the object slot instead of the more common Ner-
ven ‘nerves’ without changing the meaning (Audring & Booij, , pp.–).
() a. Du
you
gehst
go
mir
me
auf
on
die
the
Nerven.
nerves
‘You’re getting on my nerves.’
b. Du
you
gehst
go
mir
me
auf
on
den
the
Zeiger.
pointer
‘You’re getting on my nerves [lit.: pointer].’
c. Du
you
gehst
go
mir
me
auf
on
den
the
Sack.
sack
‘You’re getting on my nerves [lit.: sack].’
The fact that the ller noun phrase does not contribute anything to the meaning
of the whole utterance does not necessarily imply a high degree of productivity;
on the contrary, most lexical items that are used with this particular cxn are con-
ventionalized. However, the coercion-induced meaninglessness of the ller noun
phrase invites productivity in the sense that novel, hapax uses of the pattern (e.g.
?Du gehst mir auf den Mops [lit.: ‘pug’]) would be noticeable as non-canonical,
but decoding would not pose any problem. For the same reason, radical coercion
also invites creative and extravagant speech in the form of deliberately innovative
uses of existing patterns, at a low communicative cost. This is evident in, for
example, the lling of the noun phrase slot in the Dutch Intensifying Fake Reex-
ive Resultative Cxn (Gyselinck, ) as instantiated by [SCHRIKKEN, nom,
, ‘to be very scared’], which includes existing nouns such as hoedje ‘little
Chapter . The Devil is in the schema
hat’ in (a), but also nonsense words such as habbiebabbie in (b) (Audring &
Booij, , pp.–, ).
() a. Ik
I
schrik
frighten
me
myself
een
a
hoedje.
hat-
‘I’m very scared.’
b. Ik
I
schrik
frighten
me
myself
het
the
habbiebabbie.
habbiebabbie
‘I’m very scared.’
Instances of radical coercion are found with dierent SCxns in informal Swedish.
For example, the [fy ] cxn is found in the corpus data with (more or less) con-
ventionalized lexical substitutes such as tusan (< tusen ‘thousand’; relative fre-
quency: . tpm) as in (a), bubblan ‘the bubble’ (. tpm) as in (b), or
fåglarna ‘the birds’ (. tpm) as in (c), but also with (more or less) ad hoc
llers such as hoppsan (originally an interjection, ‘oops’; tokens) as in (a), plut-
tan ‘the little girl’ as in (b), or uttsan (a nonce word) as in (c) (the latter two
are hapaxes), in addition to the most prominent (non-taboo-avoiding) instantia-
tion fy fan (. tpm).
() a. Det
it
är
is
inte
not
kul
fun
att
to
plugga
swot
på
during
jullovet,
Christmas-break-the
fy
phew
tusan!
?thousand
‘Swotting over the Christmas break isn’t fun, hell no [lit.: phew ?thou-
[Bloggmix : avtrollis, ––]sand].’
b. Nej
no
fy
phew
bubblan
bubble-the
vad
what
trött
tired
jag
I
är.
am
‘Fuck [lit.: phew the bubble], I’m tired.’
[Bloggmix : Emma-Lous blogg , ––]
c. Vi
we
kollade
watched
på
on
2012-
2012
Och
and
fy
phew
fåglarna,
birds-the
jag
I
ville
wanted
åka
go
hem
home
‘We were watching [a movie], and fuck [lit.: phew the birds], I just
[Bloggmix : Sahra Ericzon, ––]wanted to go home.’
() a. Det
it
kan
can
verka
seem
lite
a.bit
komiskt,
funny
men
but
fy
phew
hoppsan
oops
vad
what
pinsamt
embarrassing
det
it
är
is
att
to
inte
not
känna
know
igen
again
folk.
people
‘It may seem a bit funny, but gosh [lit.: phew oops], it’s so embarrassing
[Bloggmix : Anmaja’s blogg, ––]not to recognize people.’
Steen Höder
b. Jaaa
yes
jobbat
worked
hela
whole
veckan
week-the
varit
been
helt
completely
slut
exhausted
fy
phew
pluttan
little.girl-the
‘Yes, I’d worked all week and I’d been completely exhausted, for fuck’s sake
[Bloggmix : iissiC, ––][lit.: phew the little girl].’
c. mår
feel
som
like
en
a
mumie
mummy
eer
aer
nyårsaon
New.Year’s.Eve
i
in
kroppen
body-the
men
but
fy
phew?
uttsan
vad
what
häigt
cool
och
and
kul
fun
helgen
weekend-the
var
was
‘I feel like a mummy aer New Year’s Eve, but damn [lit.: phew uttsan],
what a cool and fun weekend it was’
[Bloggmix : rackartygarna, ––]
Other SCxns are harder to identify – and in particular quantify – in a corpus
search, due to the polyfunctionality of other constructional components such as
också ‘also’ in [ också] (fan också) or wh-words such as vad ‘what’ in [.
] (vad fan), as opposed to fy ‘phew’, which is exclusively used as an interjec-
tional particle (as in fy fan). However, radical coercion can also be found with
these other cxns, as in ().
() a. Han
he
har
has
redan
already
ätit
eaten
men
but
är
is
inte
not
trött,
tired
attans
?eighteen
oxå!
also
‘He has already eaten but isn’t tired. Fuck [lit.: ?eighteen also].’
[Bloggmix : Heponimake.se, ––]
b. Snablar
[elephants’]
också!
trunks also
[Bloggmix : Biz For Real, ––]‘Damn [lit.: trunks also]!’
c. Men
but
vad
what
sjutton
seventeen
ska
shall
jag
I
köpa?
buy
‘But what the hell [lit.: what seventeen] should I buy?’
[Bloggmix : Whattowear – från mode till sjukt barn,
––]
d. Och
and
vem
who
katten
cat-the
vill
wants
bo
live
kvar
remaining
här
here
då?
then
‘And who the fuck [lit.: who the cat] would want to stay [continue to live]
[Bloggmix : Dexion, ––]here?’
Note, however, that while the polyfunctionality of constructional components
occurring in combination with taboo-avoiding lexical items poses a problem for
linguists, this need not be problematic for speakers. When decoding utterances as
in (), speakers have access to structural information that is obvious to them, but
hard to retrieve in a corpus search, such as the extraclausality and the prosodic
Chapter . The Devil is in the schema
structure of interjectional phrases (as with [ också] in (ab)) or the fact that
SWs in instantiations of [. ] (as in (cd)) are not arguments of
nearby verbs.
So far, this chapter has followed the traditional view (as expressed by e.g.
Teleman, , or Stroh-Wollin, ) in assuming that taboo-avoiding variants
are substitutes for specic SWs (e.g. katten ‘the cat’ as a substitute for fan ‘the
Devil’). From a constructional perspective, though, one could also argue that
this ultimately amounts to a quasi-derivational approach: If the lexical lling of
a SW slot in a SCxn does not contribute anything to the cxn’s meaning, then
why should, say, snablar (lit.: ‘[elephants’] trunks’) in (b) be a euphemistic
substitution for fan ‘the Devil’ rather than helvete ‘hell’? Does the SCxn specif-
ically (albeit implicitly) refer to the Devil or rather evoke some general kind
of diabolicity? While theoretical views on this question may dier, it is also an
empirical one. While a more detailed collostructional analysis (cf. Stefanowitsch,
, pp.–) would go beyond the scope of this paper, a preliminary analy-
sis based on Pearson’s chi-squared tests suggests that, for instance, [fy ] is
more strongly associated with fan ‘the Devil’ than with helvete ‘hell’ (Table ;
χ=., df = , p< .), while [ också] is not signicantly more
strongly associated with either lexical item (Table ; χ= ., df =, p= .).
Hence, [fy ] is more likely to be associated with the Devil specically, while [
också] has a more broadly diabolic meaning; yet, none of the schemas is exclu-
sively associated with one specic diabolic word.6
Table . Lexical association of fan and helvete with [fy ]
fy ¬fy total
fan , ,, ,,
helvete , , ,
total , ,, ,,
. In frequential terms, fy fan (lit.: ‘phew the Devil’) is also much more common (. tpm)
than fy helvete (lit.: ‘phew hell’; . tpm) or fy jävlar (lit.: ‘phew devils’; . tpm). As a con-
sequence, . of all instances of [fy fan|helvete|jävlar] are instances of [fy fan]. Similarly, fan
också (lit.: ‘the Devil also’) is much more frequent (. tpm) than helvete också (lit.: ‘hell also’;
. tpm) or jävlar också (lit.: ‘devils also’; . tpm), but with [fan också] only making up
. of all instances of [fan|helvete|jävlar också]. On the other hand, . of all instances of fan
occur within [fy fan], as opposed to a mere . of all helvete instances.
Steen Höder
Table . Lexical association of fan and helvete with [ också]
också ¬också total
fan , ,, ,,
helvete , , ,
total , ,, ,,
. Phonological schemas and submorphemic coercion
An additional type of constructional cue is involved in another TAS that can be
described as a subtype of lexical substitution, namely phonological modication
of SWs. This strategy is well-attested in Germanic languages. It is common in, for
example, English (cf. darn <damn,heck <hell,Gosh <God, shoot <shit), German
(verixt < verucht ‘damn’, Scheibenkleister [lit.: ‘disk-paste’] < Scheiße ‘shit’), and
Danish (søren [lit.: a male given name] < satan ‘Satan’). However, such phonolog-
ically modied SWs are usually limited to few lexicalized variants. In Swedish, in
contrast, there is a strikingly wide range of phonologically modied SWs, includ-
ing both pre-existing, phonologically similar lexical items that are used as de-
tabooed variants of established SWs (e.g. farao ‘pharaoh’, fasen ‘the phase’7<fan
‘the Devil’) and nonsense words without any literal meaning which never occur
outside swearing contexts (e.g. fasiken,fanken <fan ‘the Devil; jäklar <jävlar
‘devils, demons’; helsike <helvete ‘hell’). Moreover, such phonologically modied
forms are used frequently, and innovative forms are not uncommon.
From a constructionist perspective, such forms are interesting for three rea-
sons. Firstly, phonologically modied SWs provide additional cues that can be
used in decoding de-tabooed SW variants: For a speaker to successfully decode
a phonologically dissimilar form such as katten ‘the cat’ as a variant of fan ‘the
Devil’ requires constructional cues that are found in other components of the
SCxn (e.g. [fy ]). Phonologically similar forms, in contrast, provide an extra
cue in the form of the slot-lling SW variant itself. Secondly, as with other types
of lexical substitution, radical coercion is behind the use of phonologically mod-
. In Swedish dictionaries, fasen (as a SW) is usually listed separately from fas ‘phase’ (denite
form fasen; cf. SO s.v. fasen, SAOB s.v. fan). Even if this is accurate from an etymological per-
spective, there is no a priori reason to assume that these two lemmas are more distinct in
speakers’ mental constructicon than, say, farao used as a SW and farao used in its literal sense
(‘pharaoh’). There is, however, a prosodic distinction: Swedish is a pitch-accent language. In
tonal varieties (including most of Sweden Swedish, but excluding Finland Swedish), polysyl-
labic words (except words with ultimate stress) have either accent or accent . Fasen ‘the phase’
(accent ) is usually prosodically dierent from fasen as a SW (accent ). This is in line with a
more general tendency for SWs to have accent (cf. Footnote , p. ).
Chapter . The Devil is in the schema
ied forms, too. However, the fact that there are (some) formal constraints on
the lexical form makes them even more powerful and versatile as a resource for
extravagant language use, because the innovative forms can be used at virtually
no communicative cost, compared with the residual risk of being misunderstood
when using a dissimilar form. Thirdly, and this is a more theoretical point, modi-
ed variants constitute phonologically (partially) schematic forms.
While the phonological side of cxns in general has been somewhat neglected
in CxG, phonological schematicity has hardly been discussed in the literature at
all. Boas (, pp.–) attributes the marginal status of phonology in CxG
to the fact that the historical focus of constructionist approaches has been to
show that the lexicon, morphology, and syntax of a language form a continuum
of more or less schematic cxns, rather than being categorically distinct parts of
the language system. However, as Höder (, pp.–) argues, the exclusion
of phonology from the domain of grammar is untenable, in particular because it
ultimately has to rely on the assumption that there is a clear way of distinguish-
ing between words and other types of elements, while, in reality, the notion of
‘word’ lacks a clear denition (Haspelmath, ). So, in addition to the phono-
logical properties that have occasionally been described as parts of the form of
particular idiomatic or syntactic constructions (cf. Fried & Östman, , p.),
more general phonological schemas must be assumed for utterance-level cxns
(e.g. intonational patterns marking illocutionary force in German, cf. Höder, ,
pp.–; ingressive speech used as a back-channelling marker in Swedish, cf.
Eklund, ) as well as (morpho-)syntactic cxns, cf. Höder, , pp.–;
Riad, , p.). These arguments already suggest not only that phonologi-
cally schematic cxns are cognitively real, but also that there is no fundamen-
tal dierence between lexical and phonological schematicity. In the same vein,
phonologically schematic cxns have been proposed to account for submorphemic
patterns, such as non-concatenative morphology (cf. Höder, , pp.–;
Davis & Tsujimura, ), phonaesthemes (cf. Bergen, ; Höder, , pp.,
), and cross-linguistic sound correspondences in multilingual constructicons,
as assumed in Diasystematic Construction Grammar (Höder, ; Hagel, ).
A range of phonologically schematic types is illustrated in Table (phonological
schematicity is indicated by boldface; small-capital means any element(s); cxns
are listed in ascending order of schematicity).
As with syntactic schemas, slots can be more or less specied. For example,
the Swedish Present Tense Cxn (cf. Table ) can be formalized as [-r], which
indicates that the slot has to be lled by a verb stem (e.g. prata- ‘talk’), i.e. a very
specic morphological form. In the German Polar Question Cxn, on the other
hand, formalized as [n , rising.f], the second slot is less specied, as a
subject can take very dierent forms, such as pronouns, noun phrases, or inni-
Steen Höder
Table . Phonological schematic cxns in a schematicity continuum
Label Constructional form Example(s)
Swedish Coee [kae]kae ‘coee’
German-Danish Fine [f-{i, i}-n] German fein [fin] ‘ne’
Swedish Present Tense [-r]pratar ‘talks’
English Vision Phonaestheme [gl-]glitter,glisten,glow
Arabic Write [k--t--b]katabtu ‘I wrote’
Swedish Compounding [Accent (-)ω]bokmanus [bukmns]
‘book manuscript’
German Polar Question [n x, rising.f]Sprecht ihr Deutsch? ‘Do you
speak German?’
Swedish Back-Channelling [(, ingressive.speech)υ]ja [↓] 〈back-channelling〉
tives. Similarly, the Swedish Back-Channelling Cxn [(, ingressive.speech)υ] and
the English Vision Phonaestheme Cxn [gl-] contain slots than can be lled with
any phonological material, whereas only vowels can ll the slots in the Arabic
Write Cxn ([k--t--b]), and only two specic vowels can ll the slot
in the German-Danish Fine Cxn ([f-{i, i}-n]).
Phonologically modied SWs can be modelled in terms of phonological
schemas, too, and submorphemic coercion can thus be dened as radical coercion
involving a phonologically schematic slot, where the phonological material lling
the slot does not contribute anything to the cxn’s semantics. For the three most
common religious SWs in Swedish, the schematic SCxns in Table can be
hypothesized, including both the original SWs and the de-tabooed forms:
a. Variants of jävla(r) ‘devils, demons’ are phonological words (ω) that follow
the pattern [/j()---a(r)/ω] (shorthand: [jä--a(r)]), i.e. they consist of
disyllables starting with stressed /j()/ (long or short vowel) and ending in
/ar/ with an intermediate (phonotactically possible) consonant sequence end-
ing in a liquid (-, i.e. /-l/ or /-r/).
b. Variants of helvete ‘hell’ can be captured by the schema [/hel-σσ/ω] (short-
hand: [hel-]), i.e. they are dactylic phonological words starting with /hel/,
followed by two unstressed syllables (σσ).
c. Variants of fan ‘the Devil’ follow the pattern [/fa()-/ω] (shorthand: [fa-]),
i.e. they start with stressed /fa()/, followed by any sound or sound sequence.
The shorthand forms given here correspond to the usual (ortho-)graphic rep-
resentation of the variants.8
Chapter . The Devil is in the schema
Table . SWs and de-tabooed phonological schemas
SW De-tabooed
schema
(fully specied)
De-tabooed
schema
(simplied)
Example(s)
helvete ‘hell’ [/hel-σσ/ω] [hel-] helsike,helskota
jävla(r) ‘devils,
demons’
[/j()---a(r)/ω] [jä--a(r)] jäkla(r),jädra(r)
fan ‘the Devil’ [/fa()-/ω] [fa-] fan,fanken,farao,fasiken,
fasen
From a constructionist perspective, these schemas can be interpreted as
mother nodes within a constructional network that entails both conventionalized
and innovative, both frequent and infrequent daughters. It is worth mentioning,
though, that conventionality equals neither entrenchment nor frequency (cf. the
arguments in Schmid, , in favour of a categorical dierentiation between
entrenchment and conventionalization). For example, both the original SWs (e.g.
fan ‘the Devil’) as well as certain modied variants (e.g. fasiken) are convention-
alized, strongly entrenched in speakers’ individual grammars, and relatively fre-
quent in corpus data, whereas, say, fabylon occurs only twice in the corpus and
can be assumed to be less entrenched if at all (both tokens are found within the
same forum post by the same speaker). Besides such prototypical cases, however,
there are also instances of infrequent forms that nevertheless must be considered
lexicalized as SW variants, such as fagerlund (. tpm), which is already found
in older literary texts (Stroh-Wollin, , p.).
. Three remarks on phonetic detail: Firstly, Swedish is a pitch-accent language (cf. Footnote
, p.). As a rule of thumb, SWs – including de-tabooed variants – appear to have accent
rather than accent . Hence, the phonologically schematic SCxns should more properly be for-
malized as [/fa()-/ω], [/hel-σσ/ω], and [/j()---a(r)/ω], respectively. Hence, fasen ‘the
phase’ (accent ) is usually prosodically dierent from fasen as a SW (accent ). However, this
does not hold for all varieties of Swedish and may not hold for all variants, either (e.g. farao
can have either accent or ), and, at any rate, prosodic properties are not identiable in writ-
ten language data. As a consequence, the tonally agnostic notation with a stress mark () is pre-
ferred in this paper. – Secondly, another detail that cannot be taken into account here is the
use of a non-phonemic long front [a] (besides canonical back vowels [ ]) in fan and corre-
sponding de-tabooed variants; this phonetic contrast is not marked orthographically and can-
not be retrieved from written corpora. – Thirdly, in [/fa()-/ω], it is sucient if the rst foot
(φ) has initial stress. Hence, forms with secondary stress on the rst syllable occur as well, such
as faderullan (/faderlan/). A more precise formalization would thus be [/(fa()-)φ /ω].
Steen Höder
. Corpus evidence for productive phonological schemas
. Preliminary considerations
While the discussion in Section has demonstrated that phonologically modied
SWs can be modelled in terms of phonologically schematic SCxns, the cognitive
reality of such schemas remains an important issue. There has been much discus-
sion recently about the plausibility of highly schematic cxns (e.g. Hilpert, b;
Audring, ), against the backdrop of rich memory representations that are a
key assumption in usage-based approaches, which tend to render schemas super-
uous if their main function were to reduce redundancy within the constructional
representation (Bybee, , pp.–; Hilpert, a, pp.–). Hence, not
every possible generalization reects a schema that is cognitively real, all the less
so as there is no inherent upper limit to schematicity (Höder, , p.). How-
ever, the question whether highly schematic cxns exist can be answered empiri-
cally. One major type of evidence for their existence is productivity. In a corpus,
this can be measured in terms of a schema’s type frequency and the occurrence
of low-frequency types (Bybee, , p.; Hilpert, a, pp. –): High type
frequency of a putative schema combined with low token frequency per type indi-
cates productivity and, hence, cognitive reality.
The productivity of the assumed phonological schemas was tested quanti-
tatively using corpus data from the Bloggmix and Diskussionsforum corpora. A
number of preparatory steps were necessary for the corpus analysis:
a. Operationalization in terms of character strings: The phonological schemas
were modelled as character strings which, in turn, were transformed into
regular expressions that could be used in a CQP query. For example, [fa-]
([/fa()-/ω]) was modelled as [word=“fa.*”] (a graphic word starting with fa,
followed by zero or more characters). Since the phoneme-grapheme corre-
spondences for the phonemes occurring in the schemas are rather straight-
forward, and since spelling variation (such as äfor eor vice versa, double
letters, etc.) was taken into account, problems resulting from the interpreta-
tion of character strings as representing phonological sequences were judged
to be generally negligible.
b. Embedding in detectable patterns: Because of the polyfunctionality of SWs
(and some of their de-tabooed variants), SWs are relatively hard to identify
using a corpus search, in particular when occurring on their own. However,
some specic patterns that contain SWs turned out to be reliably detectable,
viz. the interjectional patterns [fy ] (fy fan,fy helvete,fy jävlar and substi-
tutes) and [ också] (fan också,helvete också,jävlar också and substitutes,
Chapter . The Devil is in the schema
only sentence-initial9), and the intensifying pattern [jävla ] (and substi-
tutes). These patterns were selected and translated into CQP queries. For
example, the pattern [fy fa-] was rendered as [word=“fy”] [word=“fa.*”].
c. Cleaning up raw data: The retrieved raw data were cleaned up manually.
More specically, obvious misspellings and (ortho-)graphic variants were
cautiously normalized (e.g. helveete [a typo] > helvete,jefvla [a purposely
archaic spelling] > jävla,faaaaaaan [prosodically motivated vowel redupli-
cation] > fan). Unusual substitutes that are homographs of existing lexemes
were checked manually for literal readings based on keyword-in-context
analyses; non-SW uses were excluded.
d. Assessing precision: Precision rates were calculated for each query. With the
exception of one pattern, precision was generally (close to) . For the
query used to represent the pattern [fa-också], a higher number of false-
positive results (e.g. familjer också ‘families too’, facebooksidan också ‘the
Facebook page too’) was found, even sentence-initially, because of the com-
bination of polyfunctional också ‘also’ with a catch-all phonological schema.
This resulted in a lower precision rate for this pattern (.)
The queries used are summarized in Table .
Table . Corpus queries
Schema Pattern CQP query Precision
[hel-x] [ också] [word=“h[eä]l[bcdfghjklmnpqrstvwxz]+[aeiouyäöå]+
[bcdfghjklmn
pqrstvwxz]+[aeiouyäöå]+[bcdfghjklmnpqrstvwxz]*”
& lbound(sentence)] [word=“också”]
.
[fy ] [word=“fy”]
[word=“h[eä]l[bcdfghjklmnpqrstvwxz]+[aeiouyäöå]+
[bcdfghjklmn
pqrstvwxz]+[aeiouyäöå]+[bcdfghjklmnpqrstvwxz]*”]
.
[jä--a(r)] [ också] [word=“j[eä][bcdfghjkmnpqstvwxz]+[lr]ar” &
lbound(sentence)] [word=“också”]
.
[fy ] [word=“fy”] [word=“j[eä][bcdfghjkmnpqstvwxz]+
[lr]ar”]
.
[ ] [word=“j[eä][bcdfghjkmnpqstvwxz]+[lr]a”]
[pos=“JJ”]
>.
[fa-x] [ också] [word=“fa.*” & lbound(sentence)] [word=“också”] .
[fy ] [word=“fy”] [word=“fa.*”] >.
. The search for [ också] was restricted to sentence-initial occurrences in order to exclude
as many occurrences as possible where också was used in its literal sense (‘also’), e.g. hälsoaspekt
också ‘[…] health aspect too […]’.
Steen Höder
The following sections discuss the results for the three putative schemas as
well as the observed dierences in productivity.
. Schema I: [hel-]
The hypothesized phonological schema [hel-] was tested with two patterns: [
också] and [fy ]. Tables and show the absolute and relative frequencies of
helvete ‘hell’ and its taboo-avoiding variants that occurred within these patterns.
Table . Frequencies of sentence-initial [hel- också]
Normalized form absolute relative (tpm)
helvete också .
helsike också .
helvetet också .
total ( types) .
Table . Frequencies of [fy hel-]
Normalized form absolute relative (tpm)
fy helvete , .
fy helsike .
fy helvetes .
fy helskota .
fy helvetet .
fy hällvetisk .
fy helkkari .
fy helsinki .
fy helvede .
fy helvetets .
fy helvetti .
total ( types) , .
In both patterns, the original SW is by far the most frequent slot-ller (.
tpm in [hel- också] and . tpm in [fy hel-]) and by far outnumbers all alter-
native variants combined (. tpm and . tpm, respectively). Moreover, if
we count helvete (indenite form), helvetes (indenite genitive10), helvetet (de-
. In Swedish grammaticography, ‘genitive’ is a traditional misnomer for forms with a pos-
sessive clitic -s, which has evolved from an earlier genitive singular sux. Present-day Swedish
nouns are no longer inected for case. As with the historic genitive plural form jävla, possessive
forms have sometimes lexicalized as variants of diabolic SWs, in particular when used as adjec-
tives, e.g. en helvetes explosion ‘a hell of an explosion [lit.: a hell’s explosion]’.
Chapter . The Devil is in the schema
nite form), and helvetets (denite genitive) as formal variants of the original SW,
most of the other tokens instantiate the lexicalized, conventional variants helsike
and helskota. The ve remaining tokens are hapaxes. Interestingly, most of them
are etymologically related to helvete:Hällvetisk is a (possibly purposely) archaic
spelling of helvetisk ‘hellish’, while helvede and helvetti are cognates of helvete in
Danish and Finnish, respectively; helkkari is a lexicalized Finnish taboo-avoiding
substitute for helvetti. These forms do match the [hel-] schema, but they could
also represent an alternative (or additional) taboo-avoiding strategy that involves
a deliberate distortion of lexical forms by (ortho-)graphic modication or by sub-
stitution with material from another language. For a speaker to decode helsinki,
in contrast, some kind of phonological schema must be involved. A few similar
tokens can be found in the corpora or using a web search, e.g. in () (repeated here
as ()) and (), with the variants helsinki ‘Helsinki [Finland’s capital, normally
called Helsingfors in Swedish]’ and Hälsingland (a region in Northern Sweden).
() många
many
av
of
er
you.
undrade
wondered
vad
what
i
in
helsinki
helsinki
som
.
hade
had
hänt
happened
‘many of you wondered what the hell [lit.: what in Helsinki] had happened’
[Bloggmix : My Baby Dolls, ––].
() fy
phew
Hälsingland
Hälsingland
vad
what
coolt
cool
det
it
var!
was
‘damn [lit.: phew Hälsingland], how cool was that!’
[tni.ribit.se//; accessed ––]
On the face of it, then, there is little corpus evidence for the cognitive reality of
a highly productive schema [hel-] in present-day Swedish, even though such a
schema may have played a role in the conventionalization of lexical variants in the
past. Even so, the occurrence – and assumable decodability – of very infrequent
forms suggests that such a schema is at least marginally productive. This cautious
interpretation of the data is also in line with the fact that at least small numbers
of non-conventional forms occurred with the relatively frequent pattern [fy hel-]
(. tpm), whereas no such variants were found with the overall less frequent
sentence-initial [hel- också] pattern (. tpm).
. Schema II: [jä--a(r)]
The presumed phonological schema [jä--a(r)] was tested with three patterns:
[ också], [fy ], and [ ]. Tables – show the frequencies of jävla(r)
‘devils’ and taboo-avoiding variants within the respective patterns.
Steen Höder
Table . Frequencies of sentence-initial [jä--ar också]
Normalized form absolute relative (tpm)
jävlar också .
jäklar också .
jädrar också .
total ( types) .
Table . Frequencies of [fy jä--ar]
Normalized form absolute relative (tpm)
fy jävlar .
fy jäklar .
fy jädrar .
total ( types) .
Table . Frequencies of [jä--a]
Normalized form absolute relative (tpm)
jävla , .
jäkla , .
jädra , .
jäa .
jävvla .
jäa .
jävkla .
jäckla .
jäddra .
jämra .
jäbla .
jäxla .
jästra .
jävra .
jäfra .
jäkvla .
jävbla .
jäxtra .
total ( types) , .
Chapter . The Devil is in the schema
Overall, the picture that emerges is very similar to the results for [hel-].
While the original SW (. tpm in [jä--ar också], . tpm in [fy jä--ar],
. tpm in [jä--a]) is much more common than all other variants com-
bined (. tpm, . tpm, and . tpm, respectively), most taboo-avoiding
tokens instantiate lexicalized, conventional types, with jäkla(r) being more fre-
quent than jädra(r). These are the only forms that are attested with the overall
less frequent patterns [jä--ar också] and [fy jä--ar], whereas non-conventional
forms only occur with the much more frequent pattern [jä--a]. Among those,
some can be analysed as variants of conventional ones: Jäa,jävvla, and jäa can
be considered alternative spellings of jävla (possibly indicating subtle phonolog-
ical dierences such as long vs. short vowels and voiced vs. voiceless fricatives);
similarly, jäckla and jäddra can be considered mere (ortho-)graphic variants of
jäkla and jädra. Forms like jäbla,jäkvla, or jävbla could, but need not, reect
typos. At least for jämra,jäxla,jästra,jävra,jäfra, and jäxtra ( tokens or .
tpm in total), however, it must be assumed that a (marginally productive) phono-
logical schema is involved.11
. Schema III: [fa-]
The third hypothesized schema [fa-] was tested with two patterns, [ också]
and [fy ]. Table shows the frequencies of fan ‘the Devil’ and other variants in
sentence-initial [ också].
Table . Frequencies of sentence-initial [fa- också]
Normalized form absolute relative (tpm)
fan också .
fasen också .
fasiken också .
fanken också .
fasar också .
fabian också .
fassen också .
total ( types) .
. A rarer and more conservative spelling variant containing an initial (mute) d- (djävla etc.)
was excluded from this analysis. This variant was exclusively found with the pattern [jä--a]
(, tokens or . tpm). No additional variants occurred with this spelling, and the relative
proportions of the various taboo-avoiding forms spelled with d- do not deviate substantially
from the frequencies given in Table .
Steen Höder
Again, the original SW is by far the most frequent slot-ller (. tpm), out-
numbering all the other variants combined (. tpm in total). If we interpret
fassen as a spelling variant or typo and fasar as an (unconventional) indenite
plural form of fasen, only lexicalized variants are found.
A radically dierent picture emerges, though, if we turn to the second pattern
that [fa-] was tested with, viz. [fy ]. Table shows the frequencies of fan and
other frequent variants (with relative frequencies > . tpm) in [fy fa-]. Again,
it is evident that the original SW is used much more frequently (. tpm)
than all of the other variants combined (. tpm). It is also apparent, however,
that taboo-avoiding phonologically modied variants are not infrequent either. Fy
fasen is most prominent (. tpm), followed by other relatively frequent forms
such as fy farao and fy fasiken (. and . tpm respectively) and more infre-
quent ones such as fy fanken and fy fabian (. tpm each). Unsurprisingly, these
frequent forms almost exclusively represent lexicalized, conventional variants; the
only exception, the maximally clipped form fy fa, is in most cases interpretable as
a rendering of an interrupted utterance (e.g. fy fa…!) rather than a proper lexical
substitution.
Table . Frequencies of [fy fa-] (frequent types: relative frequency > . tpm)
Normalized form absolute relative (tpm)
fy fan , .
fy fasen , .
fy farao , .
fy fasiken , .
fy fanken .
fy fabian .
fy fa .
others ( types) .
total ( types) , .
What is striking, though, is the high type frequency of [fy fa-]: Besides the
seven most frequent forms that make up the lion’s share of [fy fa-] tokens, there
are another less frequent types (given in Table ) accounting for . tpm in
total, with an average ratio of . tokens per type, including hapaxes (the aver-
age ratio for all taboo-avoiding variants is . tokens per type).
Among the less frequent types, some phonologically modied forms are lex-
icalized (such as the somewhat archaic variants fy faderullan and fy fagerlund),
whereas the majority appear to be secondary modications of more established
variants (e.g. fy fansikens is phonologically similar to fy fasiken, and fy fabikan is
Chapter . The Devil is in the schema
Table . Frequencies of [fy fa-] (infrequent types: relative frequency < . tpm)
Normalized form absolute relative
(tpm)
fy faro .
fy fasan .
fy faderullan .
fy fagerlund .
fy fader,fasikens .
fy fararo .
fy fabians,fam .
fy fad,fansiken,faskien .
fy fans,fassen .
fy fansen .
fy fanders .
fy fab,fara .
fy fallera,faran .
fy fanta, far, fasar, fav .
fy fabn,fasien,fasike,fasingen .
fy fabelen,fabikan,fabin,fablan,faderuttan,fagerdal,fana,fanan,fand,
fanden,fantomen,faren,fariken,farligt,fasa,fasikten,fasken,fassiken
.
fy faba,fabbe,fabben,fabilen,fablar,fabulan,fabylon,faderluttan,faders,
fadiken,fadren,fafa,fagerlöv,faiken,fajen,fajga,falle,falls,fankens,fankien,
fankingen,fanna,fannet,fanny,fansike,fansikens,farsken,farskon,fas,fase,
fasinken,fasisken,fasrar,fasters,fat,fatta,faun,fay
.
similar to both fy fabian and fy fasiken) or totally innovative types (e.g. fy fanny,
fy fajga). Some of the modied forms are homographs of personal names (Fabian
and Fanny are given names, Fagerlund and Fagerdal are existing family names
either derived from actual placenames or at least following a bipartite toponymi-
cal pattern commonly found in family names; alternatively, they could be inter-
preted as real or ctional toponyms).
Note that, as with the other patterns, some types could be interpreted as vari-
ants of others, if less restrictive normalization criteria were applied. For example,
fy faro and fy fararo could simply be misspellings of fy farao;fy fasa could be
a graphic variant of fy fasan or even fy fasen;fy fabbe could be derived from fy
fabian (Fabian being a given name and Fabbe a regular nickname variant). How-
ever, this cannot be decided on an objective basis. Moreover, a broader normal-
ization of [fa-] would only minimally reduce variation, while still resulting in a
vast number of infrequent types and hapaxes. Crucially, as comparison with other
Steen Höder
schemas shows (see the discussion in Sections .–.), this wide range of varia-
tion does not occur randomly with all SW schemas, but is limited to [fa-]. So,
even if we take these limitations into account, overall it remains clear that [fa-]
behaves dierently from the other schemas in that it exhibits a much higher type
frequency combined with a lower token frequency per type. This leads to the con-
clusion that [fa-], at least in [fy fa-], is a cognitively real and productive phono-
logical schema that has to be involved in the production and decoding of more
infrequent variants, while frequent types are more likely to be entrenched as cxns
of their own (cf. Figure ).
Figure . Type entrenchment vs. productive instantiations of [fy fa-]
. Dierences in productivity
The corpus analysis shows that, while phonologically schematic SCxns probably
played a role in the historical emergence of taboo-avoiding forms, their syn-
chronic productivity is more limited. However, some degree of productivity – in
terms of type and token frequencies – is observable for all three schemas that
have been analysed, with [fa-] in [fy fa-] emerging as by far the most productive
Chapter . The Devil is in the schema
one. This may in part be due to methodological choices. Firstly, fan is the most
common SW; hence, the phonological schema [fa-] is also likely to be the most
frequent of the three schemas that were analysed. Secondly, the pattern [fy ]
is more frequent than [ också]. These factors may have a positive impact on
the number of occurrences of taboo-avoiding types retrieved from the corpus, as
opposed to smaller numbers found for [fy hel-] or [fy jä--ar] or for [ också]
in general. However, comparison with the more frequent pattern [jä--a] sug-
gests that this alone does not explain the range of types found for [fy fa-], because
even if – as expected – the type frequency of [jä--a(r)] in [jä--a] is higher
than that of, for example, less common [fy hel-], the high type frequency of [fa-]
(as seen in [fy fa-]) is still exceptional.
This exception must, at least in part, be a result of arbitrary conventionaliza-
tion. However, one factor that can explain why [fa-] is particularly productive
is the observation that this cxn favours submorphemic coercion more than oth-
ers because it imposes only minimal restrictions on the phonological slot: Unlike
[jä--a(r)] and [hel-] – with their restrictions on segment types, phonotactics,
and syllable number – [fa-] can be combined with any sound material as long
as the result is a phonotactically well-formed phonological word starting with a
stressed syllable. This opens up greater potential for creative language use.
It is also worth noting that the ndings suggest that many instances of [fa-] –
conventional and entrenched ones as well as innovative forms – are organized
into a taxonomic network that connects cxns with dierent degrees of schematic-
ity, in contrast to the ‘atter’ organization of [jä--a(r)] and [hel-] variants.
Figure shows a tentative network that takes into account (some) potential
(meso-)constructions in terms of both phonological and lexical schemas, captur-
ing dierent types of formal similarity relations. Whether or not such subschemas
and networks are cognitively real is a dierent question that could, however, be
tackled empirically, in particular using experimental data in addition to corpus
data (cf. Höder, , pp. –).
This observation is also related to a more general point: Productivity is
oen dened in operational terms (cf. Bybee, , p., who denes it as “the
likelihood that a construction will apply to a new item”), and while there has
been some debate about dierent aspects of this notion (cf. Goldberg, ,
pp.–, who discusses productivity in terms of ‘coverage’, as well as Barðdal,
, pp.–, who discusses dierent subconcepts, in particular ‘extensibility’),
productivity – as measurable in, for instance, corpus analyses – is ultimately an
epiphenomenon of the way language is stored and processed in the individual
speakers’ minds (cf. de Smet, ). Contrary to traditional assumptions in gram-
mar theory, the same language – including the same structural patterns – can have
rather dierent mental representations across individuals, in particular regard-
Steen Höder
Figure . Tentative constructional network of [fa-] variants
ing the utilization of schematic as opposed to lled cxns (Dąbrowska, ). It is
likely that such individual dierences also aect the entrenchment of phonologi-
cally schematic constructions and their productivity. So, if, for instance, [fa-] can
be considered productive on the basis of a corpus analysis, this may ultimately be
based on the creative and extravagant use of this schema by few individual speak-
ers, while other speakers take up some (of the more frequent) variants, which are
then entrenched and conventionalized as lled cxns by a less creative majority of
the speech community.
. Conclusion
The relationship between swearing and linguistic taboo is, in a way, a paradoxical
symbiosis: Swearing draws its force from breaking taboos, while one of the raisons
d’être of the taboos themselves is the very possibility of breaking them. Conse-
quently, exploiting this power while manoeuvring around the problem of actually
using taboo words is obviously a relevant communicative goal. Taboo-avoiding
strategies, which are known from many languages, lend themselves to construc-
tionist research, since all techniques of euphemistic substitution of taboo elements
with innocuous ones ultimately rely on mechanisms of more or less radical coer-
cion. If semantically unrelated words or spontaneous innovations are used to
replace taboo words, then the pragmatic function of swearing has to be a property
of the schematic cxns into which the substitutes are inserted. This is all the more
Chapter . The Devil is in the schema
true if, as in Swedish, such schemas exhibit at least a minimal degree of productiv-
ity, allowing lexical slots (e.g. in [fy ]) to be lled with swearwords and conven-
tionalized de-tabooed variants as well as spontaneous innovations, which in turn
may instantiate certain sound patterns that can be modelled in terms of phono-
logically schematic cxns, such as [fa-] as a substitution for the original swear-
word fan ‘the Devil’ (e.g. fasiken,fabian,faderluttan,fanny). While this paper has
focused on Swedish, it is worth mentioning that the same mechanisms can be
assumed to be at work – although perhaps not always as extensively – in other lan-
guages as well. Although, for example, darn,shoot and heck are lexicalized, one-o
substitutions for damn,shit, and hell in today’s English, they result from phono-
logical modications of taboo words that are very similar to the ones investigated
here, and they could also be analysed in terms of – albeit less productive – phono-
logical schemas. This use of such phonological schemas in any case presupposes
the mechanism that has been labelled submorphemic coercion in this paper.
Unlike earlier centuries, religiously motivated taboos are hardly relevant any
more in today’s largely secularized society – yet speakers still make use of con-
ventionalized taboo-avoiding strategies (cf. Guardamagna, , pp.–, for a
similar observation of sociocultural determinants of productivity). Evidence for
the productivity of taboo-avoiding phonological schemas can be found in cor-
pus data, with low ratios of tokens per type pointing to the productivity and,
hence, cognitive reality of schematic cxns, albeit to dierent degrees. Whether
such variants are used in normal communication or rather jokingly, they point
to the availability of phonological schemas as resources for linguistic creativity
and extravagance. From a constructionist perspective, studying such variants can
contribute to a better understanding of phonological schematicity as well as more
general notions such as productivity, creativity, and extravagance. Moreover, what
synchronically appears as a rather marginal – indeed oen playful – utilization
of such resources also reveals a constructional mechanism that, in earlier cen-
turies, oered a useful, and potentially subversive, solution to a problem caused
by extralinguistic factors, viz. religiously motivated linguistic taboos: Phonologi-
cal schematicity may be fun, but it can also save your soul.
Acknowledgement
I wish to thank the anonymous reviewers as well as Anna Hagel and Ferdinand von Mengden
for their valuable comments on earlier versions of this paper. All remaining errors and inaccu-
racies are, of course, mine.
Steen Höder
References
Allan, K., & Burridge, K. (). Forbidden words. Taboo and the censoring of language.
Cambridge University Press.
Andersson, L.-G. (). Fult språk. Svordomar, dialekter och annat ont. Stockholm: Carlsson.
Audring, J. (). Mothers or sisters? The encoding of morphological knowledge. Word
Structure, , –.
Audring, J., & Booij, G. (). Cooperation and coercion. Linguistics, , –.
Barðdal, J. (). Productivity. Evidence from case and argument structure in Icelandic.
Amsterdam: Benjamins.
Beers Fägersten, K., & Stapleton, K. (). Introduction. Swearing research as variations on a
theme. In K. Beers Fägersten, & K. Stapleton (Eds.), Advances in Swearing Research. New
languages and new contexts (pp. –). Amsterdam: Benjamins.
Bergen, B.K. (). The psychological reality of phonaesthemes. Language, , –.
Bergs, A. (). Learn the rules like a pro, so you can break them like an artist (Picasso):
linguistic aberrancy from a constructional perspective. Zeitschri für Anglistik und
Amerikanistik, , –.
Bergs, A., & Kompa, N.A. (). Creativity within and outside the linguistic system.
Cognitive Semiotics, , .
Boas, H.C. (). Cognitive Construction Grammar. In T. Homann, & G. Trousdale (Eds.),
The Oxford handbook of construction grammar (pp. –). Oxford: Oxford University
Press.
Borin, L., Forsberg, M., & Roxendal, J. (). Korp – the corpus infrastructure of
Språkbanken. In N. Calzolari (Ed.), Proceedings of the th International Conference on
Language Resources and Evaluation, Istanbul, May –, (pp. –). European
Language Resources Association.
Bybee, J. (). Language, usage and cognition. Cambridge: Cambridge University Press.
Cappelle, B. (). What’s pragmatics doing outside constructions? In I. Depraetere, &
Raphael Salkie (Eds.), Semantics and pragmatics. Drawing a line (pp. –). Cham:
Springer.
Casas Gómez, M. (). Lexicon, discourse and cognition: terminological delimitations in
the conceptualizations of linguistic taboo. In A. Pizarro Pedraza (Ed.), Linguistic taboo
revisited. Novel insights from cognitive perspectives (pp. –). Berlin: De Gruyter
Mouton.
Culpeper, J. (). Taboo language and impoliteness. In K. Allan (Ed.), The Oxford handbook
of taboo words and language (pp. –). Oxford: Oxford University Press.
Czulo, O., Ziem, A., & Torrent, T.T. (). Beyond lexical semantics: notes on pragmatic
frames. In T. T. Torrent, C.F. Baker, O. Czulo, K. Ohara, & M. R.L. Petruck (Eds.),
Proceedings of the International FrameNet Workshop : Towards a Global, Multilingual
FrameNet (pp. –). Marseille: European Language Resources Association.
Dąbrowska, E. (). Language as a phenomenon of the third kind. Cognitive Linguistics, ,
–.
Chapter . The Devil is in the schema
Davis, S., & Tsujimura, N. (). Arabic nonconcatenative morphology in construction
morphology. In G. Booij (Ed.), The construction of words. Advances in construction
morphology (pp. –). Cham: Springer.
Diessel, H. (). The grammar network. How linguistic structure is shaped by language use.
Cambridge: Cambridge University Press.
Eklund, R. (). Pulmonic ingressive phonation. Diachronic and synchronic characteristics,
distribution and function in animal and human sound production and in human speech.
Journal of the International Phonetic Association, , –.
Evert, S. (). The IMS Open Corpus Workbench (CWB) CQP Interface and Query
Language Manual. CWB Version .. https://cwb.sourceforge.io/les/CQP_Tutorial.pdf
Fischer, K. (). Beyond the sentence. Constructions, frames and spoken interaction.
Constructions and Frames, , –.
Fischer, K. (). Conversation, Construction Grammar, and cognition. Language and
Cognition, , –.
Fjeld, R. V. (). The vocabulary of Norwegian cursing and swearing. Some of its history,
meaning and function. In M. Rathje (Ed.), Swearing in the Nordic countries. Copenhagen,
December (pp. –). København: Dansk Sprognævn.
Fried, M., & Östman. J.-O. (). Construction Grammar. A thumbnail sketch. In M. Fried,
& J.-O. Östman (Eds.), Construction grammar in a cross-language perspective (pp. –).
Amsterdam: Benjamins.
Goldberg, A.E. (). Constructions at work. The nature of generalization in language.
Oxford: Oxford University Press.
Goldberg, A.E. (). Explain me this. Creativity, competition, and the partial productivity of
constructions. Princeton: Princeton University Press.
Guardamagna, C. (). Type frequency, productivity and schematicity in the evolution of
the Latin secundum NP construction. In E. Coussé, P. Andersson, & J. Olofsson (Eds.),
Grammaticalization meets construction grammar (pp. –). Amsterdam: Benjamins.
Gyselinck, E. (). (Re)shaping the constructional network. Modeling shis and
reorganizations in the network hierarchy. In L. Sommerer, & E. Smirnova (Eds.), Nodes
and networks in diachronic construction grammar (pp. –). Amsterdam: Benjamins.
Hagel, A. (). Strange sounds, familiar words. Interlingual decoding from a CxG
perspective. Belgian Journal of Linguistics, , –.
Haspelmath, M. (). Why is grammaticalization irreversible? Linguistics, , –.
Haspelmath, M. (). The indeterminacy of word segmentation and the nature of
morphology and syntax. Folia linguistica, , –.
Hedin, C. (). Kristendomens historia i Sverige. Stockholm: Norstedt.
Hilpert, M. (a). Construction Grammar and its application to English (nd edn.).
Edinburgh: Edinburgh University Press.
Hilpert, M. (b). Higher-order schemas in morphology: What they are, how they work,
and where to nd them. Word Structure, , –.
Höder, S. (). Phonological elements and Diasystematic Construction Grammar.
Constructions and Frames, , –.
Steen Höder
Höder, S. (). Grammar is community-specic: Background and basic concepts of
Diasystematic Construction Grammar. In H.C. Boas, & S. Höder (Eds.), Constructions in
contact. Constructional perspectives on contact phenomena in Germanic languages
(pp.–). Amsterdam: Benjamins.
Höder, S. (). Phonological schematicity in multilingual constructions: a diasystematic
perspective on lexical form. Word Structure, , –.
Hoeksema, J. (). Taboo terms and their grammar. In K. Allan (Ed.), The Oxford handbook
of taboo words and language (pp. –). Oxford: Oxford University Press.
Jay, T. (). The utility and ubiquity of taboo words. Perspectives on Psychological Science, ,
–.
Ljung, M. (). Swearing. A cross-cultural linguistic study. Basingstoke: Palgrave Macmillan.
Michaelis, L.A. (). Type shiing in construction grammar. An integrated approach to
aspectual coercion. Cognitive Linguistics, , –.
Nübling, D., & Vogel, M. (). Fluchen und Schimpfen kontrastiv. Zur sexuellen,
krankheitsbasierten, skatologischen und religiösen Fluch- und Schimpfwortprototypik
im Niederländischen, Deutschen und Schwedischen. Germanistische Mitteilungen, ,
–.
Rathje, M. (). Attitudes to Danish swearwords and abusive terms in two generations. In
M. Rathje (Ed.), Swearing in the Nordic countries. Copenhagen, December
(pp.–). København: Dansk Sprognævn.
Riad, T. (). Prosodin i svenskans morfologi. Stockholm: Morfem.
SAOB = Svenska Akademiens ordbok (–) <https://svenska.se/saob; accessed
––>.
Schmid, H.-J. (). The dynamics of the linguistic system. Usage, conventionalization, and
entrenchment. Oxford: Oxford University Press.
Smet, H.de (). What predicts productivity? Theory meets individuals. Cognitive
Linguistics, , –.
SO = Sköldberg, E. (Ed.) (). Svensk ordbok utgiven av Svenska Akademien. <https://
svenska.se/so; accessed ––>.
Stefanowitsch, A. (). Corpus linguistics. A guide to the methodology. Berlin: Language
Science Press.
Stroh-Wollin, U. (). Dramernas svordomar – en lexikal och grammatisk studie i års
svensk dramatik. Uppsala: Institutionen för nordiska språk, Uppsala universitet.
Stroh-Wollin, U. (). Fula ord – eller? En enkät om attityder till svordomar och andra fula
ord. Uppsala: Uppsala universitet.
Stroh-Wollin, U. (). In the company of the Devil and Our Lord through three centuries.
Swearing in Swedish dramas. In M. Rathje (Ed.), Swearing in the Nordic countries.
Copenhagen, December (pp. –). København: Dansk Sprognævn.
Teleman, U. (). De svenska svordomarna och deras grammatik. Nysvenska studier, ,
–.
Ungerer, T., & Hartmann, S. (). Delineating extravagance. Assessing speakers’ perceptions
of imaginative constructional patterns. Belgian Journal of Linguistics, , –.
Chapter . The Devil is in the schema
