The Temperature Domain in Germanic

(1)

The Temperature Domain in Germanic

Thomas de Roo

Supervisor External advisor

Prof. Dr. P. Hendriks Prof. Dr. M. Koptjevskaja– Tamm

(Stockholm University) Second corrector

Prof. Dr. J. Hoeksema

Dissertation for the Research Master Language and Cognition

Final version

August 31st_{, 2020}

Thomas Robert Casper de Roo

University of Groningen / Rijksuniversiteit Groningen Faculty of Arts / Faculteit der Letteren

Centre for Language and Cognition Groningen

Academic Year: 2019 – 2020 Student number: s2644355

(2)

1. Introduction

The present study provides an account of the concept of TEMPERATURE1in a sample of modern Germanic languages: The West-Germanic languages Dutch, English, German, and the Scandinavian (North-Germanic) languages Swedish, Danish and Norwegian.

In daily life, much effort is dedicated to achieving and maintaining desired temperatures and to avoid ones that are uncomfortable and unsafe. Temperature is, as such, intertwined with daily life. The range of temperatures people are exposed to varies according to climate and culture, which dictates various associations with temperature. Being able to talk about these temperatures and to express the distinctions and deviations within them seems quite universal for languages.

The conceptualization of the temperature domain across languages, however, depends on a complex interaction between physical experience, external reality, and the mental representation of TEMPERATURE. The fact that temperature is so connected to subjectivity as well as cultural notions makes it an exciting subject of study.

There are many links between temperature and emotions, feelings, and other more abstracts concepts. Warmth may be used to describe something pleasant, a lovely person could have a warm heart, for example, while a distant and unfriendly person might have a cold heart. Temperature also has various connections to other senses, such as TASTE, where peppers can be hot or VISION where colors can be warm.

Temperature adjectives are hence surprisingly polysemic in many languages and can, in the right contexts, enable a variety of different readings, consider example (1) through (3), with the adjective cold:

(1) This stone is cold. [The stone feels cold on touch]

(2) John is a cold person. [John is rather unfriendly, even unsympathetic] (3) Blue is a cold color. [The color blue causes a sensation similar to cold]

(5)

Example (1) shows literal temperature, cold (2) cannot be interpreted as a physiological property of John and has to be interpreted within the domain of HOSTILITY or possibly INDIFFERENCE, while cold in (3) suggests a reading in another sensory domain: that of LIGHT/VISION.

An essential task of Lexical Typology is the task of studying how languages encode, categorize, and divide different cognitive domains among their expressions (cf. Koch 2001, Koptjevskaja-Tamm 2018). A well-known example of such research is that of the COLOR

domain (cf. Berlin & Kay, 1969). Most research within the TEMPERATUREdomain up until now

is focused on single languages or focuses on individual words and not on such lexical systems across languages families (cf. Majid & Dunn, 2015:2), on how stable they are and what differences might occur.

During the last decade, the interests of Lexical Typology shifted towards systematic analyses of cross-linguistic comparison of the lexicalization of semantic domains. A few recent examples are verbs of ROTATIONin Russian and Polish (Rakhilina, 2010), the lexicalization of theCUTand BREAK domains across Germanic (Majid et al., 2007), and verbs that express AQUA -MOTION in Germanic and Slavic (Koptjevskaja-Tamm et al., 2011), all of which reveal interesting lexical-typological differences. While it takes centuries to establish grammatical constructions, vocabulary is much more flexible: “A single generation of speakers may witness words falling in and out of use and word meanings changing dramatically.” (Rakhilina & Reznikova , 2016:101).

This study mainly draws on recently done work on cross-linguistic typological studies of the semantic domain of TEMPERATUREby Maria Koptjevskaja-Tamm (2015, ed.). This projects’ goal is to chart how various natural languages encode and conceptualize temperature and to do this systematically, according to the guidelines described in Koptjevskaja-Tamm (2007). In doing so, this work may contribute toward a typological analysis of the conceptualization of

TEMPERATURE.

The chief aim of the present study is to analyze patterns in the within-language distributions and polysemy of the lexical domain in closely related languages, contributing to cross-linguistic research by answering the following research questions:

(6)

i. What are the central temperature adjectives used across English, Dutch, German, Swedish, Norwegian, and Danish, and how similar is their distribution?

ii. How similar are the temperature systems of these closely related languages, in terms of their literal meanings?

iii. How similar are the temperature systems of these closely related languages, in terms of their possible semantic extensions?

2. Background

This study investigates the literal and abstract uses of temperature adjectives in the Germanic languages by looking at their forms and meanings. This chapter provides a general description of the TEMPERATUREdomain and how it is interpreted both literally and abstractly. Furthermore, this chapter explains the mechanisms used in temperature interpretation and how they facilitate the polysemy of temperature adjectives.

How temperature is encoded in languages is described in section 2.1. Section 2.2. explains the interpretation of literal TEMPERATUREmeanings, and section 2.3. deals with the notion of

the centrality of TEMPERATURE adjectives. Section 2.3 addresses the abstract meaning of

TEMPERATURE. In the latter three sections, examples from Dutch are taken as a point of departure.

The remainder of the chapter is devoted to the semantic extension of TEMPERATURE, seen

in the light of Cognitive Linguistics (2.5) and collocations as a means of observing the interaction between language and culture.

2.1. Temperature Values and Temperature Terms

Plank (2003) observed that “[the] experience of deviation from pleasure” lies at the basis of temperature expression values. Languages do not primarily base their temperature expressions on actual temperatures (e.g., 20° C), but instead mainly depend on the experience of thermal comfort or lack thereof, as well as temperature sensation of the surroundings (cf. Hensel, 1981:168). In both cases, TEMPERATUREis regarded relative to the average body temperature being around 37° C, and on the skin temperature being something between 32° C – 35° C.

(7)

Change in thermal comfort affects the whole body and is a reasonably subjective state faced by the experiencer. In contrast, temperature sensation is usually more objective as well as localized to a specific part of the skin being exposed to a different temperature.

Languages can differ significantly in the kind of temperature-related expressions they use in various functions. The dimension of TEMPERATURE VALUE(cf. Koptjevskaja-Tamm, 2015)

covers the distinction between warming and cooling temperatures. A TEMPERATURE TERM is any word or expression that is used to encode a particular TEMPERATURE VALUE relative to human comfort (cf. Koptjevskaja-Tamm & Rakhilina, 2006).

TEMPERATUREis often treated within the class of PHYSICAL PROPERTIES in Dixon’s list of

lexicalized concepts that often occur as adjectives (cf. Dixon, 2006) In the Germanic languages,

TEMPERATURE VALUESare primarily denoted using adjectives (cf. Plank, 2003). This thesis is restricted to adjectives denoting TEMPERATURE VALUES, used both in predicates (e.g., the winter is cold) and attributions (e.g., the cold winter), even though they exist in any word class, including verbs (e.g., to cool down), nouns (e.g., warmth), adjectives (e.g., hot) and adverbs (coldly).

TEMPERATURE VALUE Adjectives

VERY HOT benauwd ‘stuffy’

broeierig ‘sweltering’ zwoel ‘sultry’

beklemmend ‘opressive’

HOT heet ‘hot’

WARM warm ‘warm’

NEUTRAL lauw ‘lukewarm’

COOL frisjes ‘a bit chilly’

koel ‘cool/swell’ fris ‘chilly’

COLD koud ‘cold’

guur ‘bleak’ kil ‘shivery’

VERY COLD ijzig ‘icy’

Table 1 An array of Dutch temperature adjectives

Table 1 above demonstrates the diversity of Dutch temperature adjectives, many of which are rather specific and not generally applicable as modifiers. The adjective guur ‘bleak’, for

(8)

example, can only be used when it refers to surrounding temperature (see example 4), not in other cases (5-6).

Dutch

(5) *Ik drink gure thee

I drink.PRS.1SG bleak tea.

’I am drinking bleak tea’. [tactile temperature]

(6) Het weer is erg guur vandaag.

The weather COP.SG very bleak today.

‘The weather is rather bleak today’ [quasi-referential ambient temperature]

(6) *Ik heb het guur.

I have.PRS.1SG it.OBJ bleak.

‘I am bleak’ [subjective personal temperature]

Therefore, respective to the domain of TEMPERATURE, guur is somewhat peripheral. Other

temperature adjectives, such as warm warm’, are more salient and domain-central and are consequently more widely usable, consider examples (7) through (9)2.

Dutch

(7) Ik drink warme thee.

I drink.PRS.1SG warm tea.

’I am drinking hot tea’.

(8) Het weer is erg warm vandaag.

The weather COP.SG very warm/hot today.

‘The weather is rather hot today’.

(9) Ik heb het warm.

I have.PRS.1SG it.OBJ warm/hot.

‘I feel hot.’

2

Note that in the examples (7) through (9), Dutch warm is translated to English hot, rather than warm. This shows a rather interesting difference in antonymy and poses an essential question in this study (i.e., how do the temperature values relate to each other?); hot seems to be the antonym of cold in English.

In most other Germanic languages, the salient contrast is that between the encodings for warm and cold, except for Icelandic and Faroese (in both languages heitur ‘hot’ vs. kaldur ’cold’) where the contrast is as it exists in English (Remco Knooihuizen, personal communication, November 2018).

This thesis further investigates the distribution of temperature distinction and how generally usable the temperature adjectives in the major Germanic languages are.

(9)

Besides the distinction between TEMPERATURE TERMS as an expression of TEMPERATURE VALUES, there are differences in the way temperature is interpreted. Examples (7) through (9)

above are all interpreted differently. In example (7), a physical substance causes the speaker to experience a warm temperature. In (8) the temperature of the surrounding is expressed by using the weather as a placeholder for said surroundings And in (9), the speaker is expressing their take on the surroundings, i.e., they might be the only one who is hot.

The next section explains the differences in the interpretation of literal temperature meanings.

2.2. Temperature Expression and Interpretation

The domain of TEMPERATUREis described as being structured into three semantic sub-domains

(cf. Plank, 2003; Goddard & Wierzbicka 2007), as it applies to a diversity of situations and states of affairs., or FRAMES OF TEMPERATURE ELEVATION; a notion described in Koptjevskaja-Tamms 2015 volume.

The three examples (7 through 9) in the previous section correspond to TEMPERATUREbeing

evaluated within these three different frames, respectively: TACTILE, AMBIENT, andPERSONAL.

The term FRAME OF TEMPERATURE EVALUATION refers to the distinction between these three literal frames of usage and two additional abstract frames for usage outside the TEMPERATURE

domain, which is introduced in the next section.

FRAME Example Cause

TACTILE A cold stone Tangible substances/objects

AMBIENT Quasi-referential

A cold room A particular entity in the surroundings, or a type of surrounding

Non-referential It is cold. General surroundings

PERSONAL I am cold. / The dog is cold.

Anything physical affecting a person or animal’s sense of temperature

Table 2 The literal frames of temperature evaluation

When it comes to the syntax of TEMPERATUREin Germanic, the most relevant and frequent kind of expressions in this are the ones where temperature states or properties are expressed through a) predication, see example (10a) below and b) modification by attribution, see example (10b) below.

(10)

Expressing TACTILE EVALUATIONin Dutch, for example, is quite straightforward, following the structure presented above. The same constructions are used when expressing AMBIENT EVALUATION, both in quasi-referential and non-referential situations, see the examples in (11)

and (12) below.

(10) a. Attribution:

[NP (DET) [NP TempAdj N ] ]

e.g. a hot drink

DET TempAdj N

b. Predication:

[NP COP TempAdj ]

e.g. My drink is hot.

[ NP ] COP TempAdj

(11). Tactile temperature ‘hot’ a. Attribution

een het-e steen. Dutch

INDEF hot-SG.CG.STR stone ‘a hot stone’

b. Predication

De steen is heet. Dutch

DEF.CG stone be.3.SG.PRS hot.

‘the stone is hot’.

(12). Ambient temperature ‘hot’ a. Attribution

de het-e zomer Dutch

DEF.CG.SG hot-SG.CG.WK summer ‘the hot summer’

b. Predication (quasi-referential)

De zomer was heet. Dutch

DEF CG.SG summer be.3.SG.PST hot.

‘the summer was hot’ c. Predication (non-referential)

Gisteren was het heet. Dutch

Yesterday be.3.SG.PST 3.SG.N cold.

‘Yesterday, it was hot’.

The non-referential construction uses either an impersonal pronoun as subject to the copula where time or place may be given adverbially to place the predicate in context, see the cross-lingual examples in (13) below.

(11)

(13). Ambient temperature ‘cold’ Predication (non-referential)

It is cold today. English

it.3.SG.N COP cold today.

‘It is cold today’.

Vandaag is het koud. Dutch

Today COP it cold

Heute ist es kalt German

I dag är det kallt Swedish

I dag er det kaldt Norwegian

I dag er det koldt Danish

English, Norwegian, Swedish, and Danish allow for this canonical predicative construction when expressing PERSONAL TEMPERATUREas well. Standard Dutch and Standard German, by

contrast, do not allow for this construction (e.g., *ik ben warm ‘I am warm’), at least not to express personal temperature. Besides, attribution cannot be used to express personal-feeling temperature (e.g., *een warme man ‘a man who is warm’).

In Dutch, e.g., PERSONAL-FEELING TEMPERATURE is expressed using a possessive construction with hebben ‘have’ involving the presence of an empty impersonal pronoun het ‘it’ as dummy object, typical for expressing a state in Dutch (ANS:1154), see (14)a.

(14). Personal feeling ‘cold’

Ik heb het koud. Dutch

1.SG have.PRS.1.SG 3.SG.N cold

(12)

The Dutch adjectives koud ‘cold’, fris ‘cool/chilly’, and warm ‘warm’ are frame-neutral and can be used across the three literal frames, both in attribution and predication. The adjective koel ‘cool’ is non-personal and can only be used within the TACTILE and AMBIENTframes. Lauw as an intermediate NEUTRAL temperature term is very restricted and can only be used within the TACTILEframe, to express the temperature of liquids or food (e.g., een kop lauwe koffie ‘a cup

of lukewarm coffee’).

Figure 1 The hierarchy of the number ofTEMPERATURE TERMS

possible in literal temperature evaluation. (following Koptjevskaja-Tamm, 2015:19)

Koptjevskaja-Tamm (2015:19) hypotheses a hierarchy (see figure 1 above) in the number of

TEMPERATURES VALUES available to specific frames: PERSONAL-FEELING (or PERSONAL) temperature is often expressed using a reduced subset of temperature terms compared to the subset for TACTILEtemperatures, which in turn is lesser than the subset for expressing AMBIENT

temperatures. This hierarchy was already seen in effect, looking at the limited applicability of the multitude of temperature adjectives available in Dutch, in the previous section.

It seems not all temperature adjectives and the values that they encode are as salient and domain-central, or rather as basic, this is where a need for a workable definition of basic temperature terms arises. The next section discusses the notion of basic TEMPERATURE TERMS

and domain-centrality.

2.3. Basic Temperature Terms

Inspired by Berlin & Kay (1969), who defined a hierarchical system of basic COLORterms, first

Sutrop (1998, 1999) and later Plank (2003, 2010) suggested a similar system of basic

TEMPERATURE terms, cf. table 3 below. The number of basic TEMPERATURE TERMSa language can have, however, is far more limited, i.e., to 2, 3, or 4, according to Plank (2003, 2010). The

PERSONAL ≤ TACTILE ≤ AMBIENT

less more

(13)

hierarchical system of color terms, as shown in table 4 below, works in stages where the terms of each next stage only are observed in a language if the terms from the previous stage are observed.

What Plank (2003) defines as possible values in this basic TEMPERATURE system are i) WARM, ii) COLD, iii) HOT, iv) LUKEWARM, and v) ICE-COLD. A 2-term system observes only an

opposition between the values WARM and COLD, where COLD is the marked deviation from thermal comfort (which is expressed by WARM), a 3-term system adds HOT, as a marked extreme typically used for dangerous or unpleasantly warming temperatures. A 4-term system may add

LUKEWARM into the mixture, as the absence of both an unpleasant and pleasant perception.

Alternatively, instead of LUKEWARM (or NEUTRAL), the value ICE-COLDmay be observed as an elaboration on the unpleasant perception of non-warm temperatures.

Table 3. Plank’s hierarchy of temperature values (cf. 2003, 2010)

Table 4. Berlin & Kay's hierarchy of basic color terms (1969:4)

According to Sutrop, the basic temperature term is “psychologically salient, in most cases morphologically simple and native word, which generally denotes the quality of temperature at a basic level, and which is applicable in animate, inanimate and weather domains” (1998:61). This definition was criticized by Koptjevskaja-Tamm and Rakhilnia (2006) regarding his main criteria for morphological simplicity, primarily supported by evidence from Russian.

Plank’s (2003, 2010) definition of basic temperature terms is more elaborated than Sutrop’s 1998 definition; they should be (i) salient (ii) generally well-known (iii) with their meanings generally agreed on (iv). morphologically simple (v) of regular grammar (vi) native or at any

2-value → 3-value → 4-value

Warming ↑ ↓ Cooling WARM COLD + HOT NEUTRAL or ICE-COLD

Stage I → Stage II → Stage III/IV → Stage V → Stage VI → Stage VII

BLACK WHITE + RED + GREEN and/or YELLOW + BLUE + BROWN + PURPLE PINK ORANGE GRAY

(14)

rate nativized (vii) specialized for the TEMPERATURE domain or primarily used for the

TEMPERATURE domain (viii) within TEMPERATURE hardly restricted in their application (cf.

Plank 2010).

According to the definition above, not all Dutch adjectives introduced in section 2.1. are central temperature terms. Koud and warm are the most salient temperature adjectives coming to mind when asking any Dutch speaker, and it is the main distinction used. Heet ‘hot’ and koel ‘cool’ are options for the third basic temperature term in Dutch, of which Plank's definition prefers heet. At this point, the two options for the fourth basic temperature, according to Plank's hierarchy, that are left are lauw ‘lukewarm’ and ijskoud ’ice-cold’, and this poses a problem.

(15) a. The coffee is lukewarm. b. *The stones are lukewarm.3

The term lauw and its equivalents in other Germanic languages (e.g., English: tepid, German lau, Swedish ljummen) do not directly come to mind and are therefore not salient (requirement i; Plank 2003, 2010). Neither are they general in their application (viii; Plank 2003, 2010) because their usage is mainly limited to the TACTILEframe, and even within that mainly only to liquids and possibly food, see example (15) above.

Whether or not the Dutch term ijskoud is salient, is debatable. It might not be among the first temperature terms coming to mind, but it is much more salient than lauw. It is also not as restricted as lauw is, i.e., for Dutch, it can be used among all three literal frames and has few restrictions in its combinability with nouns. It is, however, not non-compositional or at any rate morphologically simple (Planks’ requirement iv), i.e., it consists of the morphemes ijs- ‘ice’ and koud ‘cold’. IJskoud can hence never be a basic temperature term, because of this morphology being a part of an elaborate system of adjective intensification which exists in Germanic.

To express intensification or modification of the value expressed by a temperature adjective, i.e., raising and lowering of its value, Dutch (and many other Germanic languages)

(15)

makes not only use of regular adverbs of degree or standard intensifiers (e.g., heel warm ‘very warm’, verschrikkelijk koud ‘awfully cold’) but also of elative compounding (cf. Hoeksema, 2012). In elative compounding, the left-member indicates a high degree of the property expressed by the right-member of the compound, e.g., smoorheet ‘smothering hot’. The elevating compound adds additional semantic information, such as the source of the temperature, the entity through which it is experienced, or how it is experienced.

In ijskoud, elative compounding is used, the left-member ijs- adding semantical information about the similarity to ice. As such, ijskoud is not suitable as a basic temperature term. However, saying that Germanic uses a 3-value system is reductive, at least. Other salient temperature adjectives directly come to mind when thinking of temperature, but do not correspond to any of the values in Plank's hierarchy. For Dutch, those are koel and fris, as we already saw in section 2.2. above.

Plank’s (2010) two options of the four-value system are not attested in the languages touched upon in the 2015 volume of Koptjevskaja-Tamm (2015:29-30). The described languages in that volume with four value system typically distinguish between ‘cold’, ‘cool’, ‘warm’, and ‘hot’ (ibidem). Therefore, I consider the Germanic four-value system to consist of the four values COLD,COOL,WARM, and HOT, with the addition of a marginal NEUTRALvalue. Plank (2010) adds a reservation about LUKE (in his terms), that whenever there is a term to

express neutral temperatures, it is unclear to what extent this truly is a basic term, generally being restricted to food and weather.

TEMPERATURE TERM Number of occurrences Relative frequency

heet 66 465 8.65 % warm 345 839 45.00 % lauw/lauwwarm 11 561 1.50 % fris 117 660 15.31 % koel 33 195 4.32 % koud 193 737 25.22 % Totals 768 457 100 %

Table 5. The total frequencies of the domain central temperature adjectives in the nlTenTen Corpus 20144

(16)

The total frequencies of the supposed central terms in Dutch, as found in the nlTenTen corpus (Jakubíček et al., 2013), are shown in Table 5, above. These data includes the non-literal usages. It shows that warm ‘warm’ with 45% and koud with 25.22% are by far the most common temperature terms, which makes sense as it is the main temperature distinction made in Dutch, e.g., warme en koude dranken ‘hot and cold drinks’ and koude en warme landen ‘cold and hot countries’.

The central temperature distinction in English is between hot and cold, e.g., hot and cold drinks and hot and cold meals. In English, hot spreads from pleasantly warming to unpleasant temperatures and even dangerous ones, whereas in Dutch heet ‘hot’ is much more marginal.

The third most common term fris ‘cool/chilly’, with 15.31%, is mainly used to express ambient temperature, e.g., Het is fris vandaag ‘It is chilly, today’. The fourth most common term heet refers to unpleasant or possibly dangerous temperatures, e.g., Kijk uit, hete koffie! ‘Watch out, hot coffee!’. With 4.32%, the adjective koel ‘cool’ and with 1.50 % the intermediate adjective lauw ‘lukewarm’ are the least frequent.

There is no strict border between 3- and 4-value systems as suggested by Koptjevskaja-Tamm (2015:29); “It is in general unclear whether the ‘cool’ terms ever become as frequent as those for ‘cold’, ‘warm’ and ‘hot’ even when they may apply to all the three frames of temperature evaluation.”. The analysis of this study compares the frequency proportions of these four central values and the marginal NEUTRAL value across a subset of the Germanic

languages to decide on their centrality.

All literal uses of the TEMPERATUREdomain we have seen up until now can be described using the three frames explained in the previous section. Temperature adjectives are actually surprisingly polysemic and can enable a variety of different readings away from temperature-proper; these examples are repeated as (16) and (17) below.

(16) John is a cold person. [John is rather unfriendly, even unsympathetic] (17) Blue is a cold color. [The color blue causes a sensation similar to cold]

(17)

Considering these again, they are expressed using the same attributive construction as when a literal reading is implied. However, their meaning does not directly have anything to do with the temperature value associated and is only supported by similarities between the literal and the abstract meaning. The next section elaborates on such extensions of meaning.

2.4. Extended Meaning

FRAME Example Usage Experience

EXTENDED The man was killed in cold blood.

Metaphorical/metonymical usages, mapping

temperature onto domains outside the temperature, and even outside the perception domains Any abstract experience with some cognitive similarity to temperature.

CROSS-MODAL The chili peppers are

hot.

Mapping temperature concepts to other

perceptual modalities (such as taste, vision, audition, olfaction)

Anything present in the real world that is perceived similar to temperature. Table 6 The two frames outside the temperature domain

In addition to the three literal FRAMES OF TEMPERATURE EVALUATION introduced in, two additional frames can be distinguished in which TEMPERATUREis evaluated abstractly, these are clarified in table 6 above5.

Unlike the literal FRAMES OF TEMPERATURE EVALUATION (TACTILE, AMBIENT, and PERSONAL), these abstract frames differ in that they generally show little to no variation in the syntactic constructions they are used in. The construction used for the TACTILE frame could be considered the most unmarked in syntax (i.e., the default).

At this point, it is reasonable to mention that the syntax typical to the TACTILE frame,

especially the canonical predicative construction, is not specific to TEMPERATUREalone, and is

used across concepts as a general syntax for expressing physical properties (cf. Dixon, 2006). This notion makes it possible to theorize that TEMPERATUREas a rather physical concept is the basis of extended interpretation. While the attributive and predicative syntax is the same as for

5_{I decided upon these in consultation with Prof. Dr. M. Koptjevskaja-Tamm (personal communication,}

(18)

TACTILE TEMPERATURE usage, the combination of the temperature adjective and the noun prompts a different reading.

Direct Indirect

TACTILE CROSS-MODAL PERSONAL AMBIENT EXTENDED

CROSS-MODAL

Figure 2. A scale of the directness of the five frames of temperature evaluation

The semantic difference between the two abstract frames lies in the extent to which the concepts are perceivable in the real world (directness. EXTENDED TEMPERATURE EVALUATION covers a wide range of abstract topics, from feelings (e.g., a warm feeling) to manifestations of human responsiveness (e.g., a cold heart; someone with a cold heart is not likely to respond kindly).

CROSS-MODAL TEMPERATURE EVALUATION covers topics and objects that are present in the real

world or physically experienced in some way or another, e.g., a cold color is something that actually can be noticed by the human eye. Interpretation within the CROSS-MODALframe can be considered both more or less on the “direct”-side of the scale, depending on the modified noun, e.g., a warm sound denotes something more ambient than a hot pepper, which is physically present when eating it. Figure 2 above illustrates the relationship between these frames and the real world.

Table 7 above lists some corpus examples to illustrate the difference between these two frames. As we can see, the ABSTRACT examples are directly perceivable in the real world. They might, however, cause physiological events or changes associated with temperature, which are perceivable, this holds for example for AFFECTION, which is associated with warmth (cf. Lakoff

& Johnson, 1980, Kopjevskaja-Tamm, 2015).

CROSS-MODAL

COLOR cool color

EXTENDED

AFFECTION warm hug VISION cool light INDIFFERENCE cold heart

SOUND hot rhythm RELEVANCE hot subject

TASTE hot salsa INTENSITY hot discussion

(19)

TOUCH chilly sensation

Table 7. A selection of examples (translated to English) of CROSS-MODAL and ABSTRACT TEMPERATURE EVALUATIONin Germanic Languages6

For abstract interpretation, there is some cognitive connection with TEMPERATURE, which results in an indirect correspondence between expressions and cognitive structures.

Consequently, there ought to be some mechanism in our cognition that maps TEMPERATURE

upon other semantic domains. In such cases of semantic extension, Cognitive Linguistics helps us understand how the human mind works in deducting, interpreting, and producing meaning. 2.5. Theory of Meaning

As soon as we see or hear form, we assume meaning. A simple form can prompt the construction of an extremely complicated meaning (cf. Fauconnier and Turner 2003:65). We are not entirely conscious of the mental processes that get us this meaning, primarily when the sum of the parts does not correspond with the whole intended meaning.

Cognitive Linguistics, primarily founded in the late 1970s and early 1980s by Lakoff, Talmy, and Langacker, provides us with theoretical devices useful in the analysis of the rich polysemy observed in the evaluation and interpretation of temperature adjectives in various contexts, especially with meaning outside the TEMPERATUREdomain.

2.5.1. Meaning and Embodiment

Within Cognitive Linguistics, meaning is considered to be encyclopedic (cf. Croft, 1993). There is no defined clear border between ‘linguistic meaning’ and any order kind of knowledge. Language is purely understood to be a means of expressing cognitive structures. It is assumed that there is no essential difference between cognitive representations of linguistic knowledge and those of other (general) knowledge (Croft, 1993:337).

Paradis (2001:48) mentions that the “activation of conceptual patterns in the cognitive system” makes meaning arise from linguistic utterances, i.e., there can be a direct correspondence between certain linguistic expressions and specific cognitive structures. Deane

(20)

(1988:325) points at the perspectival nature of linguistic expressions and mentions the natural consequences of the human ability to think flexibly: semantic contrast, polysemy, and indeterminacy in meaning.

Thoughts are abstract representations of the real world, grounded and based on repeated experience patterns: regularities experienced using aspects of reality perceived (VISION,SOUND, SPACE), or executed by the body (MOTION,FORCE) (cf. Langacker, 2008:28-32). It makes sense that the concepts expressed in language are ultimately grounded in bodily experience, even from a developmental point of view (cf. Tomasello, 1992; MachWhiney, 1999).

Embodiment, as an idea stemming from the fields of physical anthropology and developmental psychology, is the notion that substance and spirit are bound and that human psychology arises from the body’s physiology. Cognitive Linguistics implies that meaning is embodied, such that the speaker’s bodily experience influences what linguistic expressions are used to carry meaning to the hearer (cf. Evan & Green, 2006).

2.5.2. Metaphors and Metonymy

As can be taken from the classic work by Lakoff and Johnson (1980), the metaphor is a crucial device in understanding the way humans conceptualize things (including emotions and feelings) in the mind. Metaphors emerge by domain mapping, i.e., expressing a something from one lexical domain in terms of another domain, which is often made possible by certain cognitive correlations. Such mappings are based on the similarity of at least one aspect of both the source and the target domain. Metaphors are therefore expressed as equations, as seen in (18) below.

(18). [TARGET DOMAIN] IS [SOURCE DOMAIN]

AFFECTION IS WARMTH

e.g., warm hug

RELEVANCY IS HEAT

e.g., hot issue

INDIFFERENCE IS COLD

(21)

Embodied experience (cf. § 2.3.2.) is a viable source domain in the emergence of such correlation-based conceptual metaphors (cf. Lakoff & Johnson, 1980; Kövecses, 2010). Kövecses mentions that at the base of such conceptual metaphors such as INTENSITY IS HEATlie

“bodily correlations in experience” (2015:6): as the intensity of states or activities increases, the production of body heat increases as well, which is an unavoidable aspect of having a body. The concepts of ANGER,LOVE,LUST,WORK, andARGUMENT all have a property of INTENSITY, which allows for mapping towards the correlation-based metaphors such as ANGER IS HEAT,

LOVE IS HEAT(ibidem).

Another fundamental theoretical tool in Cognitive Linguistics is metonymy, which is also

important for the mapping between domains. Traditionally, metonymy is defined as using one

entity to refer to the other (cf. Taylor, 1989:122). In contrast, in Cognitive Linguistics, it is defined as a collection of mental processes that shift meaning from one entity to another within the same conceptual domain (Croft, 2003: 177). Metonymy allows for highlighting target content to replace entities within the same conceptual domain, in result backgrounding source content, i.e., to focus on the RESULTrather than the CAUSE(Cosghignano, 2019:20-21).

Such metaphors as INTENSITY IS HEATare made possible by the metonymical process of

meaning shift. At the base of this shift lies the metonymic operation PHYSIOLOGICAL EFFECT FOR EMOTION, a specialization of a CAUSE FOR EFFECTmetonymic operation.Intense emotions

or actions cause, e.g., an increase in body heat and are as such, using a metonymic mapping. The RESULTbody heat can be used to express the CAUSE.

Metonymic mapping works recursively so that in effect, anything mapped on e.g. the target INTENSITYcan be mapped upon e.g. HEAT, resulting in a broad array of possible

metaphoric extensions, supporting the polysemy as seen in temperature terms.

2.6. Adjectives and Collocations

In the tradition of the Moscow School of Semantics as well as Corpus Linguistics (cf. Firth, 1957; Wierzbicka, 1985; Sinclair, 1991; Apresjan, 2000; Stubbs, 2001; Kilgarrriff & Tugwell, 2002) the meaning of words is considered to be dependent on context. Words have the potential to activate different cognitive structures based on in which context they appear.

(22)

Adjectives are abstract and vague when their meaning is considered in isolation. In context, their meaning is decided upon, or rather, influenced by the nouns that they modify, as well as other information available in the broader context of the particular discourse.

The combinatory potential of adjectives with specific nouns (i.e., the types of nouns an adjective can be combined with) is thus rather essential for grasping its meaning; in this, I follow Stathi (2015:360), who researched the combinability of Greek temperature adjectives. According to this approach, the meaning of a word is formed by its context, rather than its lexical denotation (Koptjevskaja-Tamm & Rakhilina, 2006).The Distributional Hypothesis, inspired by Structural Linguistics and the distributional methodology of Zellig Harris (1968), cited in Koptjevskaja-Tamm & Sahlgren, (2014:238) hypothesizes a correlation between similarity of distribution and similarity in meaning. This hypothesis is tested by the corpus-driven part of this study, comparing the meaning of cognates in closely related languages.

The concept of a collocations was first established by Firth (1957:11-14) as a denotation of the syntagmatic relationship between particular words, in phrases or fragments where the selection and order of words is not free. Schönefeld (2007:139) considers collocations in a slightly more extensive way, including “not only […] a word’s preferences for the company of particular other words, but also for particular syntactic categories they attract” (2007:138). Collocations range from idioms to fragments with variable items, which “co-occur more often than chance would predict” (Schönefeld, 2007:137).

As mentioned in section 2.3, language is a means to express cognitive structures, which in turn are influenced by what a person already knows about reality, i.e., by the mental models they made to understand the world around them. For a full picture, distributional information needs to be considered together with cognitive approaches to meaning acquisition.

Usage data about collocations reflecting specific cultural notions or models can thus help us understand the “conventional mental models” (Schönefeld, 2007:139), which can be utilized to account for “what people talk about and how they do it” (ibid.).

Collocations are a rather useful tool, to find out what i) the central temperature terms are, ii) how similar they are in their literal meanings, and iii) how similar they are in their possible semantic extensions are. As collocations serve as a window into the underlying cognitive structures that are recruited in interpreting the polysemy of temperature across different

(23)

semantic domains, I expect to find suggestions of deviation in the underlying conceptualizations which are associated with the temperature terms in the six languages I research.

3. Methodology

In the following chapter, the method and materials used in this study of the literal and abstract meanings of Germanic temperature adjectives are presented. In the first section, the corpora used and the database I built for data analysis and visualization are introduced, and in the second section, the data set used in the further analysis is presented.

The analysis of the distribution is presented in chapter 4. The findings, in terms of meaning, are presented in chapter 5.

3.1. Data Collection

3.1.1. Corpora and Language Selection

For the corpus-driven analysis, it is desirable to have corpora that are as extensive and as comparable as possible. The TenTen corpus family (Jakubíček et al., 2013) was found to comply with those requirements and is a collection of comparable corpora of web text, freely available for research through SketchEngine. The web texts are crawled from the internet and contain news articles, forum posts, openly accessible social media and blogs, as well as more formal texts found on the internet. Therefore, they are very representative of actual language usage in formal and informal situations.

The initial data collection, based on elicitation and surveys by using the temperature guidelines provided by Koptjevskaja-Tamm (2007), included survey data for Icelandic, Faroese, Afrikaans, Yiddish, Swiss-German, and Low-Saxon (Gronings variety), collected by Koptjevskaja-Tamm and me. As the corpus material that is available for these languages is very sparse, it was decided not to include these languages in the corpus-driven methodology of this thesis.

To ensure a fair comparison concerning the distributions of TEMPERATURE FRAMES, VALUES and the distinctions between them, the choice was made to include the six major

Germanic Languages: English, Dutch, German, Swedish, Norwegian and Danish. The reason for making this choice is the relevancy of using corpus material as comparable as possible.

(24)

Table 8 below shows an overview of these six corpora in the TenTen corpus family and their respective sizes in number of tokens.

Corpus name Year Corpus language Corpus size (in tokens)

deTenTen13 2013 German 16.526.335.416

enTenTen15 2015 English 15.703.895.409

svTenTen14 2014 Swedish 3.401.035.817

nlTenTen14 2014 Dutch 2.538.714.434

noTenTen17 2017 Norwegian (bokmål) 2.472.622.031

daTenTen17 2017 Danish 2.170.994.053

Table 8 An overview of the six corpora used in this study (cf. Jakubíček et al., 2013)

3.1.2. Collocation Database

I created the online published Collocation Database of Germanic Temperature Adjectives (cf. De Roo, 2019), henceforth referred to as CDGTA. This database charts the literal (tactile, ambient and personal) and abstract (extended and cross-modal) uses of temperature adjectives, by containing an extensive collection of nouns found in collocations with temperature adjectives.

For each language, the 150 most frequent attributive and the 150 most frequent predicative collocations for each encoding (TEMPERATURE TERM) of the four basic TEMPERATURE VALUES

(HOT,WARM,COOLandCOLD) and the marginal value NEUTRALwere selected.

The TEMPERATURE TERMS included in the database were chosen either based on earlier

research (for Swedish (cf. Koptjevskaja-Tamm, 2006), English (cf. Rasulic 2015 & Schönefeld 2007) and German (cf. Schönefeld 2007)), or on the data collected according to Koptjevskaja-Tamm’s guidelines for temperature data collection (for Dutch, Norwegian and Danish) and comply to Plank’s (cf. 2003, 2010) definition of basic temperature terms, see § 2.1.1. Table 9 below gives an overview of the included adjectives. For some TEMPERATURE VALUES

synonymous terms or other salient terms were chosen as well (e.g., chilly and cool for the cool value in English).

(25)

COLD koud cold kalt kold kald kall

COOL koel, fris cool, chilly kühl,

frisch

kjølig, kølig, sval

kjølig kylig, sval

NEUTRAL lauw lauwwarm lukewarm, tepid lau, lauwarm

lunken lunken ljum,

ljummen

WARM warm warm warm varm varm varm

HOT heet hot heiß hed het het

Table 9 The temperature adjectives used in collecting data

The database was populated with the most frequent combinations of temperature adjectives with nouns, in attributive and predicative syntax using SketchEngine’s useful WordSketch feature. WordSketch summarizes an individual word’s behavior based on corpus data, see figure (4) below. It shows e.g., nouns that an adjective is likely to modify as well as other words that co-occur with it.

Figure 4. A screenshot of SketchEngine’s WordSketch of the English adjective ‘warm’

The frequencies for all attributive collocations were gathered using WordSketch, as this data was readily available through SketchEngine for all six corpora. For the predicative constructions (the verbal constructions where a copula is combined with a temperature

(26)

adjective) however, only for English, Dutch, and Norwegian WordSketch data was readily available. For the other three languages, additional queries were needed to gather those, these queries were written to approach the data collection done by WordSketch itself and are given in Appendix B. They resemble the underlying structure given in (19)a below. The general structure in (19)b represents the attributive construction entirely collected through WordSketch.

(19). a. [ NP ] [ COP ] [Adj]

Example: The recent weather was warm

b. [NP (DET) [Adj] [N] ]

Example: (a) hot drink

For the sake of overview, Table 10 below shows for which languages and constructions WordSketches or alternative constructions were used.

Language Corpus Predication Attribution

Dutch nlTenTen WordSketch data

English enTenTen

Norwegian noTenTen

Danish daTenTen Additional

queries in Appendix B

Swedish svTenTen

German deTenTen

Table 10. An overview of the methods used to gather the most frequent collocations

The CDGTA consists of the database and a web tool I made to browse through the collocations. A typical entry looks like Figure 6 below. It contains information about the collocation: the type (temperature term and modified noun) and the number of tokens (absolute hits), tags for the FRAME OF TEMPERATURE EVALUATION (AMBIENT,TACTILE,PERSONAL,EXTENDED orCROSS

(27)

-MODAL) and a semantic category (such as water, surfaces, time) as well as additional distributional information.

Figure 3 A typical entry in the CDGTA (cf. De Roo (2019) accessed October 23rd₂₀₁₉₎7

3.1.3. Tagging the collocations

The CDGTA contains 8778 collocations with temperature adjectives, both as attributes and predicates, of which the attributive part (5355) was semantically tagged manually by myself. The predicative part was mainly tagged automatically, based on the earlier tags of the attributive part. These automated annotations were manually checked and corrected where needed.

The collocations were tagged according to their meaning, using a list of semantic tags compiled for this task, in consultation with Koptjevskaja-Tamm (personal communication, October 2018). The categories used are based on Koptjevskaja-Tamm (2007), Schönefeld (2007), Kövecses (1995), Geeraerts & Grondelaers (1995), also Goossens (1998) and Lakoff & Johnson 1997:50 and draw on Koptjevskaja-Tamms notion of TEMPERATURE ENTITIES (cf.

2007, 2011).

The process of tagging consisted of extensive corpus searches, careful inspection of what a term means in various samples in the context, as well as consultation with native speakers. Some annotations required extra attention and discussion to decide on the most appropriate

(28)

category. During the process, this led to some reconsideration in which some of the categories were merged with others, split into several, or simply omitted.

In the analysis, I reconsidered and merged some categories into more manageable ones without losing the general semantic information contained in them. The final list of semantic tags I used in my analysis is presented below in table 11. Appendix A contains the original list of tags used in the CDGTA and how they translate to the tags used in the analysis of this thesis.

TACTILE

Household and consumption Nature Liquids Human body AMBIENT Time Place

Weather and environment Indoors

Bed and clothing Sources and conductors

PERSONAL Subjectiveness CROSS-MODAL Tactile perception Audative perception Olfactory perception Visual perception EXTENDED Affection Calmth

Intensity and danger Indifference and logic Relevance and topicality Passion and lust

Popularity and extraordinarity Table 11. An overview of semantic tags used in the semantic tagging

At this point, it is reasonable to mention that while tagging, some collocations were found to be polysemous, and as such, were given multiple tags. In the database, their absolute hits were divided equally by the number of tags they were given (in which every single tag is seen as a database entry), this was done to approach their possible distribution and avoid overrepresentation. This means that the number of collocations for a certain semantic tag does not need to be an integer.

(29)

3.2. Data sample

The variables present used in the analysis, and their respective number of values are: language (6), temperature value (5), syntactic construction (2) and frame of evaluation (5). This yields a possible 300 combinations and makes analyzing the data of the CDGTA a complex task. For this reason, some of the factors needed to be merged for the analysis to be executable and manageable.

Initial analysis of the data considered how attributive and predicative collocations differ in the within-language distributions of the TEMPERATURE VALUES. For attributive constructions,

the overall (both the literal and abstract collocations) differences in the distributions of

TEMPERATURE VALUE were found to be significant, X 2 (20, N = 5 763) = 132,49; p < .001. However, in predicative constructions there were not any significant differences in distribution, X 2(20, N = 2 982) = 5,78; p = 1. Appendix C gives an overview of the data used in these tests. As the predicative constructions did not contribute to any interesting differences, I decided to merge the factor syntactic construction in further analysis.

Table 12 below gives an overview of the FRAMES OF TEMPERATURE EVALUATION that are

distinguished in each of the three analyses.

To answer research questions i), ii) and iii), which consider the distribution of and similarity of temperature terms in their literal (ii) and abstract (iii) meanings, it suffices to distinguish between semantically literal usages of TEMPERATURE and semantically abstract usages of TEMPERATURE. For this purpose, the collocations that were tagged to belong to the literal FRAMES OF TEMPERATURE EVALUATION, being TACTILE, PERSONAL, and AMBIENT, were combined into a single group of semantically literal usage. The same was done to the EXTENDED

and CROSS-MODAL FRAMES OF EVALUATION, they were placed in the semantically abstract

group.

To answer the research question i), when it comes to the centrality of temperature terms, careful inspection of the three literal frames (TACTILE,AMBIENT andPERSONAL) is needed to decide on the domain centrality of the terms. To address this domain-centrality the distributions of the terms across the three original literal FRAMES are used, as well as the four ‘candidate’

central temperature values: HOT,WARM,COOL and COLD. AsNEUTRAL is too marginal to be

(30)

Analysis Frames included Temperature values Merged Research question LITERAL ABSTRACT Literal meaning distributions

TACTILE AMBIENT PERSONAL

CROSS -MODAL EXTENDED HOT WARM NEUTRAL COOL COLD yes, into LITERAL i), ii) x x x

Factors: language, merged frame, temperature value

Abstract meaning distribution

CROSS -MODAL EXTENDED x x HOT WARM NEUTRAL COOL COLD yes, into

ABSTRACT i), iii)

Factors: language, merged frame, temperature value

Domain centrality

CROSS -MODAL EXTENDED x x x HOT WARM COOL COLD no; TACTILE, AMBIENT, and PERSONAL i)

Factors: language, original literal frame, temperature value Table 12. The factors used in the three analyses of this thesis

The final data sample is broken down by language and TEMPERATURE TERMfor each of the three

analyses in tables 13, 14, and 15 below.

English Dutch German

HOT 445,965.5 36.62 % 30.056 11.64 % 431 781 22.87 % WARM 282,710 23.21 % 121.409 47.02 % 463 938.5 24.57 % NEUTRAL 6,353 0.52 % 6.873.5 2.66 % 569 90.5 3.02 % COOL 129,757 10.66 % 29.356 11.37 % 447 846.5 23.72 % COLD 353,014 28,99 % 70.538 27.32 % 487 647.5 25.83 % Total 1.217 799.50 100% 258.233 100% 1 888 204 100%

Swedish Norwegian Danish

HOT 13 516 5.21 % 3 312 1.32 % 2 348 0.87 %

WARM 136 035 52.43 % 127 337 50.93 % 147 729 54.86 %

NEUTRAL 8 117 3.13 % 5 347 2.14 % 6 602.5 2.45 %

COOL 18 994.67 7.32 % 13 105.5 5.24 % 17 913.5 6.65 %

(31)

Total 259 456.2 100% 250 041 100% 269 286.5 100%

Table 13. The tokens per language and TEMPERATURE TERM for the literal collocations.

Percentages given are within language and frame.

HOT 195 007 36.43 % 7 790 10.88 % 133 106 41.11 % WARM 114 385 21.37 % 45 996 64.24 % 90 961 28.09 % NEUTRAL 5 874 1.10 % 3001 4.19 % 2 965 0.92 % COOL 162 838 30.42 % 9 518 13.29 % 35 154 10.86 % COLD 57 147 10.68 % 7 992 11.16 % 61 577 19.02 % Total 535 251 100% 71 597 100% 323 763 100%

HOT 26 142 37.25 % 7 777 18.82 % 5 989 10.16 % WARM 28 550 40.68 % 16 066 38.87 % 34 581 58.64 % NEUTRAL 564 0.80 % 1103 2.67 % 777 1.32 % COOL 2 762 3.94 % 2223 5.38 % 3 632 6.16 % COLD 12 159 17.33 % 14 159 34.26 % 13 991 23.73 % Total 70 177 100% 41 328 100% 58 970 100%

Table 14. The tokens per language and TEMPERATURE TERM for the abstractly tagged

collocations. Percentages given are within language and frame.

Language Frame Value N % Language Frame Value N %

English ᴘᴇʀsᴏɴᴀʟ ʜᴏᴛ 443 15.03% Swedish ᴘᴇʀsᴏɴᴀʟ ᴡᴀʀᴍ 597 16.40% English ᴘᴇʀsᴏɴᴀʟ ᴡᴀʀᴍ 152 5.16% Swedish ᴘᴇʀsᴏɴᴀʟ ᴄᴏᴏʟ 14 0.38% English ᴘᴇʀsᴏɴᴀʟ ᴄᴏᴏʟ 94 3.19% Swedish ᴘᴇʀsᴏɴᴀʟ ᴄᴏʟᴅ 3017 82.86% English ᴘᴇʀsᴏɴᴀʟ ᴄᴏʟᴅ 2259 76.63% Swedish ᴛᴀᴄᴛɪʟᴇ ʜᴏᴛ 7391 7.36% English ᴛᴀᴄᴛɪʟᴇ ʜᴏᴛ 201999 46.91% Swedish ᴛᴀᴄᴛɪʟᴇ ᴡᴀʀᴍ 50079 49.86% English ᴛᴀᴄᴛɪʟᴇ ᴡᴀʀᴍ 6218.5 14.44% Swedish ᴛᴀᴄᴛɪʟᴇ ᴄᴏᴏʟ 2291 2.28% English ᴛᴀᴄᴛɪʟᴇ ᴄᴏᴏʟ 26554 6.17% Swedish ᴛᴀᴄᴛɪʟᴇ ᴄᴏʟᴅ 40671.5 40.50% English ᴛᴀᴄᴛɪʟᴇ ᴄᴏʟᴅ 139875 32.48% Swedish ᴀᴍʙɪᴇɴᴛ ʜᴏᴛ 6112 4.15% English ᴀᴍʙɪᴇɴᴛ ʜᴏᴛ 24523.5 31.31% Swedish ᴀᴍʙɪᴇɴᴛ ᴡᴀʀᴍ 85359 57.96% English ᴀᴍʙɪᴇɴᴛ ᴡᴀʀᴍ 220376.5 28.33% Swedish ᴀᴍʙɪᴇɴᴛ ᴄᴏᴏʟ 16689.67 11.33% English ᴀᴍʙɪᴇɴᴛ ᴄᴏᴏʟ 103109 13.25% Norwegian ᴀᴍʙɪᴇɴᴛ ᴄᴏʟᴅ 39105 26.55% English ᴀᴍʙɪᴇɴᴛ ᴄᴏʟᴅ 210880 27.11% Norwegian ᴘᴇʀsᴏɴᴀʟ ᴡᴀʀᴍ 123 9.92% Dutch ᴘᴇʀsᴏɴᴀʟ ᴄᴏʟᴅ 1385 100% Norwegian ᴘᴇʀsᴏɴᴀʟ ᴄᴏᴏʟ 12.5 1.01% Dutch ᴛᴀᴄᴛɪʟᴇ ʜᴏᴛ 14743 10.62% Norwegian ᴘᴇʀsᴏɴᴀʟ ᴄᴏʟᴅ 1105 89.08% Dutch ᴛᴀᴄᴛɪʟᴇ ᴡᴀʀᴍ 77028 55.50% Norwegian ᴛᴀᴄᴛɪʟᴇ ʜᴏᴛ 2428 1.77% Dutch ᴛᴀᴄᴛɪʟᴇ ᴄᴏᴏʟ 8366 6.03% Norwegian ᴛᴀᴄᴛɪʟᴇ ᴡᴀʀᴍ 93392 68.03% Dutch ᴛᴀᴄᴛɪʟᴇ ᴄᴏʟᴅ 38664 27.86% Norwegian ᴛᴀᴄᴛɪʟᴇ ᴄᴏᴏʟ 1481.5 1.08% Dutch ᴀᴍʙɪᴇɴᴛ ʜᴏᴛ 15313.5 13.77% Norwegian ᴛᴀᴄᴛɪʟᴇ ᴄᴏʟᴅ 39969.5 29.12% Dutch ᴀᴍʙɪᴇɴᴛ ᴡᴀʀᴍ 44381 39.92% Norwegian ᴀᴍʙɪᴇɴᴛ ʜᴏᴛ 884 0.83%

(32)

Dutch ᴀᴍʙɪᴇɴᴛ ᴄᴏᴏʟ 20990 18.88% Norwegian ᴀᴍʙɪᴇɴᴛ ᴡᴀʀᴍ 33822 31.85% Dutch ᴀᴍʙɪᴇɴᴛ ᴄᴏʟᴅ 30489 27.42% Norwegian ᴀᴍʙɪᴇɴᴛ ᴄᴏᴏʟ 11614.5 10.94% German ᴘᴇʀsᴏɴᴀʟ ʜᴏᴛ 864 9.90% Danish ᴀᴍʙɪᴇɴᴛ ᴄᴏʟᴅ 197 56.38% German ᴘᴇʀsᴏɴᴀʟ ᴡᴀʀᴍ 737 8.45% Danish ᴘᴇʀsᴏɴᴀʟ ᴡᴀʀᴍ 113.5 15.48% German ᴘᴇʀsᴏɴᴀʟ ᴄᴏᴏʟ 1789 20.51% Danish ᴘᴇʀsᴏɴᴀʟ ᴄᴏᴏʟ 113.5 8.92% German ᴘᴇʀsᴏɴᴀʟ ᴄᴏʟᴅ 5333 61.14% Danish ᴘᴇʀsᴏɴᴀʟ ᴄᴏʟᴅ 962.5 75.61% German ᴛᴀᴄᴛɪʟᴇ ʜᴏᴛ 31716 34.57% Danish ᴛᴀᴄᴛɪʟᴇ ʜᴏᴛ 845 0.57% German ᴛᴀᴄᴛɪʟᴇ ᴡᴀʀᴍ 213274.5 23.24% Danish ᴛᴀᴄᴛɪʟᴇ ᴡᴀʀᴍ 88744 59.78% German ᴛᴀᴄᴛɪʟᴇ ᴄᴏᴏʟ 158461 17.27% Danish ᴛᴀᴄᴛɪʟᴇ ᴄᴏᴏʟ 5504 3.71% German ᴛᴀᴄᴛɪʟᴇ ᴄᴏʟᴅ 228661 24.92% Danish ᴛᴀᴄᴛɪʟᴇ ᴄᴏʟᴅ 53346 35.94% German ᴀᴍʙɪᴇɴᴛ ʜᴏᴛ 113756 12.57% Danish ᴀᴍʙɪᴇɴᴛ ʜᴏᴛ 1503 1.33% German ᴀᴍʙɪᴇɴᴛ ᴡᴀʀᴍ 249927 27.62% Danish ᴀᴍʙɪᴇɴᴛ ᴡᴀʀᴍ 58788 52.04% German ᴀᴍʙɪᴇɴᴛ ᴄᴏᴏʟ 287596.5 31.78% Danish ᴀᴍʙɪᴇɴᴛ ᴄᴏᴏʟ 12296 10.88% German ᴀᴍʙɪᴇɴᴛ ᴄᴏʟᴅ 253653.5 28.03% Danish ᴀᴍʙɪᴇɴᴛ ᴄᴏʟᴅ 40385 35.75% Swedish ᴘᴇʀsᴏɴᴀʟ ʜᴏᴛ 13 0.36%

Table 15 The tokens per language and TEMPERATURE TERM for the original literal frames.

Percentages given are within language and frame.

4. Distribution of the temperature adjectives

This chapter focuses on the quantitative part of the analysis of temperature collocations. It provides an overview of the distribution of the four central TEMPERATURE VALUES, with the addition of the NEUTRAL value, in the six major Germanic languages, when used with literal and abstract meanings.

(33)

Figure 4. An overview of the distribution of temperature values across the six languages, divided by

semantic usage

Figure 4 above gives an overview of the mean relative frequency of TEMPERATURE VALUES

represented by the TEMPERATURE TERMS (cf. 3.1.2.) that encode them, in two different conditions based on the expression of literal meanings (TACTILE,AMBIENTand PERSONALtaken together) and abstract meanings (EXTENDEDand CROSS-MODALtaken together), this makes for

a total of 60 combinations in figure 4. The distributions are given as a proportion of the sums of the mean relative frequency (per 1 million words). The frequencies associated with this figure are given in overview tables 14 and 15 below.

For the mean frequencies of the within-languages distributions, bootstrapped means and confidence intervals (95%) were computed using resampling in R; these are given in Appendix I.

Chi-square analyses on the within-language distributions for the literal (X2 _{(20, N = 6 203)} = 127,29; p < .001) and abstract (X 2 (20, N = 2 542) = 127,29; p < .001) cases, revealed significant differences, elaborated on further in this chapter.

HOT 2858.75 47.16 % 297.59 17.85 % 2 260.63 29.42 % WARM 1262.10 20.82 % 714.17 42.85 % 1 893.63 24.64 % NEUTRAL 84.71 1.40 % 85.92 5.16 % 134.73 1.75 % COOL 477.05 7.87 % 161.30 9.68 % 1 468.78 19.11 % COLD 1378.96 22.75 % 407.73 24.46 % 1 927.46 25.08 % Total 6061.57 100 % 1666.71 100 % 7685.23 100 %

HOT 125.15 11.84 % 87.16 7.58 % 26.68 2.23 % WARM 531.39 50.26 % 511.39 44.49 % 674.56 56.41 % NEUTRAL 25.61 2.42 % 93.81 8.16 % 63.49 5.31 % COOL 46.67 4.41 % 65.87 5.73 % 62.63 5.24 % COLD 328.55 31.07 % 391.23 34.04 % 368.46 30.81 % Total 1057.37 100 % 1149.46 100 % 1195.82 100 %

(34)

In the following sections, these distributions are addressed, starting with the literal temperature meanings in section 4.2 followed by the distributions across the semantically abstract cases in section 4.3.

Section 4.4 revisits the question of central temperature terms and makes use of the distributional data to determine to what extent the hypotheses about the centrality of temperature terms hold for these six languages. After breaking down the collocations by meaning in the next chapter, this question will be once more addressed.

HOT 1500.05 35.82 % 116.26 7.27 % 1 331.06 25.52 % WARM 1299.83 31.04 % 901.87 56.36 % 2 115.37 40.55 % NEUTRAL 35.60 0.85 % 11.13 0.70 % 20.88 0.40 % COOL 579.49 13.84 % 190.35 11.90 % 540.83 10.37 % COLD 772.26 18.44 % 380.57 23.78 % 1 208.39 23.16 % Total 4187.23 100 % 1600.18 100 % 5216.53 100 %

HOT 142.08 16.29 % 60.76 8.52 % 53.00 4.82 % WARM 483.90 55.47 % 382.52 53.67 % 720.44 65.54 % NEUTRAL 9.25 1.06 % 22.06 3.09 % 17.65 1.61 % COOL 23.81 2.73 % 29.63 4.16 % 39.05 3.55 % COLD 213.31 24.45 % 217.82 30.56 % 269.06 24.48 % Total 872.35 100 % 712.79 100 % 1099.2 100 %

Table 18. Mean absolute frequencies per TEMPERATURE VALUEin abstract usage.

4.1. Literal Meaning Distributions

Looking at the differences between the relative frequencies of TEMPERATURE VALUES used

literally, it becomes clear that for all languages except English and German, WARMand COLD

form the main temperature opposition based on frequency.

While the German frequency of literal HOT is significantly higher than in the other languages, it fails in opposition to COLD, at least in the more tactile cases. In all languages but English, the TACTILEand AMBIENTusages of HOTsolely involve some aspect of DANGER or at

the very least INTENSITY, cf. section 5.2.

As stated in section 1, the English adjective hot spreads from pleasantly warming to unpleasant temperatures and even dangerous ones. The main temperature contrast in English is that between COLDand HOT. This contrast is best illustrated in attestations in conventionalized pairs (Koptjevskaja-Tamm, in press, 18), such as meals.

(35)

English is the only of the six languages in which the adjective hot indeed behaves as the adjectives encoding the value WARM do in the other languages, i.e., expressing pleasant and

desired warming temperatures. Compare the attestations of collocations with ‘meal’ and semantically related nouns (whenever ‘meal’ was not attested) and their PMI as calculated by SketchEngine (cf. Kilgarriff et al., 2014) in table 19 below, the higher the PMI score, the more likely the combination is to occur. Note that the combination with WARMdoes occur in English, but with a much lower PMI-score of 5.86 vs. 2.35 for hot vs. warm + meal, which makes sense as hot and cold meals (cold and hot meals) is the more fixed and conventionalized phrase, e.g., to be found on restaurant menus.

COLD WARM HOT

English meal 1.18 meal 2.36 meal 5.87 Dutch ‘buffet’ 2.07 ‘buffet’ 2.66 German ‘buffet’ 1.28 ‘buffet’ 1.27 Swedish ‘meal’ 0.92 ‘meal’ 1.73 Norwegian ‘dinner’ 4.90 ‘dinner’ 2.41 Danish ‘refreshment’ 2.11 ‘meal’ 0.68

Table 19. Attested collocations with meal and related nouns and their PMI-scores

Looking at co-occurring adjectives in conjunctions with an adjective is another feature of SketchEngine’s WordSketch (cf. Section 3), and tells us something about the distinction between two adjectives. Table 20 below shows that cold co-occurring with hot happens predominantly in the English corpus, and much less or even not at all in the other five languages.

A Chi-square analysis revealed that the association between the frequency of co-occurrences of WARM/COLDand HOT/COLDand the variable language is very significant in the case of English, Dutch, Swedish, Norwegian, and Danish. X2(4, N = 51 558) = 31 505, p < 0.001. This means that the co-occurrence of HOT and COLD has a higher chance to occur in English than in the other four languages. For German, no data about co-occurrence was available. The effect size of this association between language and co-occurrence is rather large

(36)

(Cramer’s V = .782). Therefore, the fact that English has opposes cold and hot more than the other languages is not due to chance.

‘cold and/or warm’

‘warm and/or cold’ ‘cold and/or hot’‘cold and/or hot’

Per million Absolute hits Per million Absolute hits

English enTenTen 0.32 5 806 1.57 28 938

Dutch nlTenTen 1.42 3 703 0.04 100

German deTenTen Unavailable

Swedish svTenTen 0.16 609 < 0.01 13

Norwegian (Bokmål) noTenTen 2.83 8 213 0 0

Danish daTenTen 1.63 4 189 0 0

Table 20 Co-occurrences of cold and warm vs. cold and hot8

As can be seen in table 20, English has a higher amount of co-occurrences of hot and cold together than the other four languages (German excluded). makes it reasonable to accept hot as the salient antonym of cold in English, while for the other for languages it is clear that (in the case of co-occurrences) warm and cold is the preferable opposition.

When it comes to the distributions of individual TEMPERATURE VALUES across

languages, the most considerable significant differences between pairs of languages exist within

HOT (with a large effect size (r =.228), see table 21 below) and NEUTRAL (with a large effect

size (r = .208), see table 17).

As expected, in all of the six languages, the adjectives for the value NEUTRALhave the lowest collocation frequency and are thus very marginal. The Scandinavian trio of languages have a modestly higher share of NEUTRAL terms(see table 17 above), by comparison with any

language except Dutch. When broken down in semantic tags (cf. 5.1.), all languages except Swedish seem to use NEUTRAL mainly for TACTILE temperature expression. The usage of

NEUTRAL is predominantly with liquids, i.e., tepid water, and is attested in all six languages9. The effect size of the different distributions of WARMis found to be moderate (r= .064, see table 21), mainly because apart from in English, where there are significant differences with all other languages (see table 21), the frequencies of WARMare about the same. Few significant

differences between the other five languages occur. The same holds for COLD, with an

8_{The co-occurrence feature of WordSketch was unavailable for the German deTenTen corpus.} 9_cf._{http://tderoo.nl/gertemp/overview/domain/neutral}

The Temperature Domain in Germanic