Expressing smells in (American) English

4.2.1 Collexemes of smell

The data of smell were more numerous than the ones of the other smell verbs. That is why, it was possible to carry out collexeme analyses for most of the smell constructions in the different periods. The earliest corpus did not contain enough data for all the constructions to make collexeme analysis reasonable or informative, except for ‘smell + N/obj’.

For the other smell verbs, the smaller amount of data available suggests that collexeme analyses be carried out for fewer patterns, namely, ‘scent + N’ (all periods), ‘scent with’ (2nd and 3rd period), ‘reek with’ (1st and 2nd period), ‘reek of’ (2nd and 3rd period), ‘stink + adj’, ‘stink like’ and ‘stink of’ (all period 3).

In the following, we will discuss the transitive pattern ‘smell + N’ in more detail, with the remaining patterns, for reasons of space, being ‘reduced’ to mentioning their peculiarities.

Table 3 gives you the list of the top 30 most strongly attracted collexemes for the three periods of usage covered by the data.

The data are informative about the usage of this pattern in the respective time periods.

In general, the collexemes allow for the identification of different senses of the transitive pattern, namely

1. that of ‘perceiving something by the sense of smell’ (cf. sense vi) in Table 1 above), suggested by smell brimstone, sulphur, incense, fragrance, for example,

2. that of ‘suspecting something (on the basis of alleged evidence)’, suggested by smell a rat and smell a thief.

   Another sense (listed as sense i) in Table 1) is

3. that of ‘exercising the sense of smell’.

From the collexemes of the verb, the latter sense cannot be differentiated from the first, because we do not see the subjects of these constructions. These would allow for the identification of the type of event expressed, a perception of something (sense (i)) or an action of using one’s sense of smell (sense (iii)).^[22] That means the subjects would have had to be identified in order to distinguish between the two senses. However, for reasons of time, the necessary data were not (yet) determined and explored.

The identification of more than one sense of the pattern poses the question if we analyze several constructions (as form-function pairings) rather than just one, a question raised more generally by Gries (2023a). He points to the problem that often alleged constructions are reduced to their formal side, “without taking the functional side into consideration (enough)”. To remedy the situation for our data, we would indeed have to manually re-adjust the input data to the computation, something planned for the next phase of our project. Furthermore, the amount of data extracted from the subcorpora, especially the earlier ones, do not seem to us sufficient for a more stringent classification in this respect, so that we consider the extension of the database as a possibility to carry out such a more ‘fine-grained’ analysis.

Looking in more detail at how people talk about the perception of smell, we investigated the collexemes of the individual periods for what they say about the ‘things’ speakers draw on for the description of smells. In light of what researchers have found about the scarcity of specific words codifying smells (see Section 2), it is of special interest to identify and classify the ‘things’ employed as good mediators of the kind of smells people experience and want to communicate.

The nouns complementing the verb smell in the transitive pattern usually name the thing from which the smell emanates rather than explicitly describing a particular quality of the smell. In other words, the fillers found are ‘source-based terms’ (cf. Croijmans and Majid 2016), giving us ‘source-based’ descriptions of smells (cf. Poulton 2023: 294). In such uses, the quality and the respective evaluation of the smell as pleasant or unpleasant is implied in the given source term, as a result of speakers’ knowledge of and their experiences with the smell-emitting objects named.

This suggests that the collexemes are also informative about the sources of smells that are salient enough in a culture and (well-)known enough in the respective speech community to be used as functional smell descriptors. Poulton (2023) aims at discovering more about smell sources commonly invoked by English speakers, analyzing the nouns around the verb smell as used in the TV/MOVIES genre of the COCA (Davies 2010). This is also an aim of our own study, which, however uses a different database, wants to trace potential changes in speakers’ usage, and looks at 3 other smell verbs as well. Still, we will build on some of the results Poulton reports, in particular with respect to the categories of filler words he used and identified.^[23] The top 90 of the lemmatized nouns he found (ordered by total frequency in the corpus) could be assigned to 11 conceptual domains (cf. Poulton 2023: 302) and we will see how the (top 30) collexemes identified here (ordered by LLR values) fit into these categories, or are of a different type.

Table 4 gives a survey of what we observed for the transitive patterns of the verbs smell and scent.^[24]

Looking at the verb smell, we can see that the fillers fit into 4–7 of Poulton’s 11 domains. This lower number may be due to the much smaller number of items we considered,^[25] and to our zooming in on more specific patterns and on postverbal nouns only. On the other hand, our data required that we add some other categories to Poulton’s repertoire, namely

1. metaphorical uses of smell (as in sb smelled a rat/fear), mainly referring to some (mostly negative) situation assumed or sensed by its user (which were excluded from Poulton’s analysis),^[26]

2. ‘cognate object’ uses (for example, sb smelled a smell/ an odor), providing very general information on the perception of a noticeable (mostly negatively evaluated) smell, and

3. unspecified object uses (sb smelled it), in which the smell descriptors (such as thing, elements, something, or that, it) are too general to classify them, or cannot be read off the collexemes themselves.^[27]

Moreover, we noticed that for some fillers, there is a certain degree of ambiguity in the domain assignment, because they fit into more than one domain, e.g. stockyard, fitting into both ‘food and farming’ and ‘architecture, buildings, houses and home’. In such cases, we opted for only one domain, for instance, the more ‘literal’ one or the one semantically closer to ‘something that emits a smell’ (the former in the example).

Finally, and most importantly, unlike Poulton’s (2023: 312) list, our filler lists result from the computation of strongly attracted collexemes. That is why they represent not just the most frequent fillers, but the ones most strongly attracted to the pattern. So, the strongest collexemes and their rankings can be taken to hint at the domains most commonly invoked in ‘smell talk’ when using the verb smell in the transitive pattern. The arrangement of domains in Table 4 reflects the collexeme ranking and can be inspected for the importance of the respective domains and potential changes over time.^[28] If we also consider the type numbers of the individual domains as an indicator of prominence in smell talk, the domains of ‘substances, materials & objects’ and ‘body & individual’, and ‘food & farming’ stand out in smell + N. In metaphorical uses, smell a rat can be found in the data of all periods, and cognate object uses show an overlap, too (e.g. smell an odor). The pattern ‘scent + N’ covers quite some metaphorical uses, mainly in the 2nd period, and further attracts the domains of ‘world & environment’ and ‘social’.

Taking a diachronic perspective, there are some changes to be observed in the ranking and the domains invoked. For ‘smell + N’, the domain ‘life & living things’ is not represented in the data of the first period, that of ‘social’ only appears in the data of the second period, and ‘world & environment’ is missing from the data of the 3rd period.^[29]

Within the domains identified, some collexemes have noticeable token frequencies, especially in the 3rd period (e.g. smoke = 31 in period 3, and 17 in period 2). The type-token ratios (TTRs) (for the top ranks) reflect this situation (1st period 0.94, 2nd – 0.27 and 3rd – 0.09). Though a decrease in the TTR as such speaks for a lower degree of lexical variation in the data, we must, however, bear in mind a general increase in the number of instantiations of the transitive pattern over time. In the later periods, we find larger type numbers and larger token numbers at the higher ranks of the filler words. That means that, over time, smell descriptions seem to be given more often, and the fillers employed are becoming more ‘entrenched’, but simultaneously more diverse. Hence, it is also worth looking at the top collexemes themselves, because increases in their token frequencies may allow for the assumption that some lexically filled patterns are becoming more common (such as smell smoke, or perfume), and, conversely, those patterns decreasing in token number (such as smell brimstone or fragrance) may be understood to ‘fall out of use’. From such findings we can infer changes in American English speakers’ preferred ways of communicating smells when they use the verb smell in a transitive pattern. In fact, we can trace an increase in the use of odorous substances (domain of body & individual) as fillers: smell a scent/perfume/cologne/aftershave in period 3, whereas period 1 only has ‘smell a fragrance’, and period 2 – perfume and fragrance. This may have to do with the emergence and development of a cosmetics industry, promoting people’s use of such substances, so that their smells are more generally known and possibly considered worth communicating.

Looking at the more negatively evaluated filler words, we find that they can be assigned to the domains of ‘substances, materials and objects’, ‘food & farming’ and ‘body & individual.’ Period 1 exhibits the use of brimstone, sulphur (substances), blood (body), as well as smell and odor (as cognate objects), period 2 – that of smoke, gas, sulphur (substances), manure (farming), breath (body), grub (food), and odor, and period 3 – that of smoke, something burning (substances), blood, sweat, breath (body), and odor and stench (as cognate objects). Also here, the fillers diversify over time. Assumingly reflecting an increasing and more conscious awareness of the body, the domain of ‘body & individual’ is getting more broadly exploited. The lists also include an increasing number of offensive words, perhaps as a reflection of a generally less restricted or more liberal use of taboo words in the 20th century.

4.2.2 Collexemes of scent

As said above, collexeme analyses were carried out for ‘scent + N’ and ‘scent like N’ only. For these patterns, the top 30 collexemes (if there were that many) were analogously assigned to their conceptual domains (see Table 4 above), with the results being a bit different from the transitive smell pattern. There is a noticeable increase in the occurrence of metaphorical uses in period 2 (total of 32 tokens and 19 filler types, vs. 3 in the first and 2 in the 3rd period). Moreover, the domain of ‘world & environment’ (fillers: air, gale, wind) is represented in the fillers of all three periods, ranking on top in the 2nd and 3rd. The fillers allocated to this domain brought up an interesting problem. This has to do with the function of the filler nouns in the communication of smell. Most of them serve the function of expressing the thing whose smell is perceived and communicated by the speaker, such as sb smells bread, rose, smoke. The fillers of this domain, however, may also refer to the thing that is the carrier of the/a smell, as in (flowers) scent a table , or steakhouse (scented by smoke). Poulton (2023: 298f) differentiates these functions as ‘stimulus’ and ‘descriptum’ in a smell experience.^[30] In such cases, the subjects in the utterances play an important role in the disambiguation: flowers scented the table, or smoke scented the steakhouse make the post verbal noun a descriptum, whereas a dog scented the steakhouse would have the steakhouse as stimulus. This situation poses the same general problem that had become apparent in the ‘smell + N’ patterns: The pattern is associated with more than one meaning, and would, more precisely, represent two constructions. And, as for ‘smell + N’ (cf. the previous section), we would need to consider the subjects of the patterns in order to assign the respective hits to the one or the other construction. These have not (yet) been extracted from the concordances. Once this is done, the data can be investigated by means of a co-varying collexeme analysis and the different senses can be isolated from the output data more easily. However, in view of the comparatively small database (21/85/28 hits in the three subcorpora), it is suggested that it should be extended for the application of this method.

As for the actual filler words, two of them turn up in Winter’s (2019) investigation of English words and the sensory modalities they can express. For the olfactory sense, he reports only two example nouns, namely air and breath (cf. Winter 2019: 199), which he sees as an indication of a low lexical differentiation of the smell sense (cf. Winter 2019: 222). In our data, air is strongly attracted to the scent + N pattern (rank 1 in subcorpora 2 and 3, and rank 13 in subcorpus 1). In the smell + N pattern it ranks 19th in the 3rd subcorpus, but does not show up in the top 30 ranks of the 2nd subcorpus and not at all in the first one. (smell) breath is absent from the first subcorpus, but ranks 25th in the 2nd and 11th in the third one. In the top collexemes of scent, it does not occur at all. That means that these two words are much less employed in smell talk than Winter’s reported observation would make us expect. Instead, our data suggest that speakers have incorporated into smell talk a growing repertoire of other words.

The nominals following scented occasionally invite for an adjectival reading of the past participle verb form, as in scented wedding invitations, scented soap. However, the concordances show that these object nominals occur as subjects of a passivized transitive construction and were classified accordingly, i.e. as objects of a transitive construction.

The diachronic development of the transitive scent construction is also different from that of smell in that the TTRs are more even in the 3 subcorpora (0.86, 0.60 and 0.79, respectively) and we can observe the highest frequency of usage in the 2nd period, with a decrease in usage frequency in the 3rd subcorpus. The numbers must, however, be interpreted with caution because of the different sizes of the corpora.

We will now turn briefly to what the respective collexeme lists reveal about the use of the four smell verbs in the predicative patterns with prepositional phrases (aka complex intransitives) ‘smell verb + of N/ like N/ with N’.

4.2.3 Collexemes of smell verbs complemented by prepositional phrases

Because of the small numbers of hits extracted for some of these patterns in the different subcorpora, only a subgroup thereof was submitted to a collexeme analysis. Tables A and B in the Appendix give an overview of the domains the fillers could be assigned to.

The pattern ‘smell like N’ was investigated for subcorpora 2 and 3 (Table A), ‘smell of N’ for all three of them (Table B).

The first pattern most strongly attracts fillers from the domains of ‘substances’ (short for ‘substances, materials & objects’) (subcorpus 2) and ‘body’ (short for ‘body & individual’) (subcorpus 3). Perhaps the latter finding (subsuming such fillers as shit, sweat and fart) can also be understood as an indication of changes in the use of taboo language already addressed in Section 4.2.1. The domain of ‘substances’ attracts 5 different fillers in each period, with the fillers in the 2nd subcorpus (e.g. something scorching and incense) ranking more highly. There is also a substantial number of fillers from the domain of ‘food’ (short for ‘food & farming’) (10 in the 2nd, such as doughnut and onion; 6 in the 3rd period, such as lemon and meat). ‘Life’ (short for ‘life & living things’) is represented by 6 fillers in the 2nd period (such as isthagaria and herring), and, finally, metaphorical usage^[31] is rare in both periods.

The second pattern’s top-ranking fillers align with the same domains, though in different orders: the most strongly attracted filler in subcorpus 1 (camphor) also aligns with the domain of ‘substances’, but the most strongly attracted filler in subcorpus 2 (bay rum) – with ‘food’, and that of the 3rd subcorpus (sweat) – with ‘body’. All these domains show decent type numbers (from 3 to 18) and there is a general increase in types to be noticed over time, speaking for a diversification in the ‘smell vocabulary’.

Comparing this, also with the transitive pattern discussed above, it is not surprising to find that the ‘body’ domain is more prominent in the ‘smell of N’ pattern. For, a smell caused by some aspect of one’s body or its functions can be seen to be in a ‘part of’ relation with one’s body, motivating the use of this preposition.

The collexemes in the ‘scent with N’ pattern (calculated for the 2nd and 3rd periods) are mainly associated with the domains of ‘food’ (such as spice (period 2) and mango (period 3)) and ‘life’ (such as flower (both periods) and lavender (period 3)), with an increase in the type numbers in the 3rd period.

‘stink like N’ (3rd subcorpus) has been found to mainly attract fillers from the domains of ‘substances’ (absinthe, sewage) and ‘life’ (skunk, cow). The high TTR of 1.0 speaks for lexical variation in this construction rather than the emergence of lexically filled patterns. ‘stink of N’ is more frequent, and, as a consequence, has more filler types with more tokens. They come from, above all, the domains of ‘body’ (sweat, urine) and ‘substances’ (antiseptic, animal waste). There are, however, also quite a few metaphorical uses, as in stink of illiteracy, condescension. All the examples, inclusive of the latter, reflect the general association of the verb stink with negatively evaluated smells, so that the verb seems to be equipped with a negative semantic prosody.

The ‘reek with N’ pattern (2nd period) shows a surprisingly high number of metaphorical uses (reek with hypocrisy, corruption), all of which express negatively evaluated phenomena. A second domain prominent in this pattern is ‘substances’ (chloroform, grease), where the fillers refer to things that are less generally considered negative (liniment, spirits). And there are fillers from the ‘body’ domain (toilet water, cologne), which – as such – are positively ‘loaded’ in our modern world. It can, however, be assumed that the evaluations are turned more negative by the verb reek, which seems to have acquired a negative semantic prosody, too. This assumption is supported by the few hits from period 1 (with such fillers as blood, smoke, fever). The 7 hits (in this period) with the filler blood, are conspicuous in that they suggest that speakers at that time had it available as the lexically filled pattern reek with blood. Moreover, it could have acted as a sort of ‘mother construction’ for further uses of ‘reek with N’.

The ‘reek of N’ pattern is a bit more diversified with respect to the filler nouns found: they associate with 8 domains in the 2nd period data, where ‘substances’ are most strongly attracted, and with 6 domains in the 3rd period data, where ‘substances’, ‘body’ and ‘food’ are noticeable in the higher ranks. The fillers from the domain of ‘food’ as such beef stew, fried food, do not refer to smells generally perceived as negative, but again, the verb’s negative semantic prosody may depict the experience of the smell as (more) negative in the respective situation. Metaphorical uses, more prominent in 3rd period, are also concerned with such properties and behaviors that are held in lower esteem in the eyes of society.

4.2.4 Collexemes of smell verbs complemented by adjectives

This use was classified as an instantiation of the predicative construction (cf. Table 1 above). Because of the small amount of hits extracted from our database, collexeme analyses were only carried out for the adjectives following the verbs smell (2nd and 3rd subcorpus) and stink (3rd subcorpus). Table C in the Appendix gives the results.

The adjectives identified in the smell pattern contain just one of those that had been defined as a basic smell word (cf Section 2): musty. It is fairly strongly attracted to the pattern, ranking 3rd in the second subcorpus and ranking 6th in the most recent subcorpus. From the adjectives listed by Winter (2019: 197, cf. Section 2.) as infrequent for the olfactory sense, burned showed up in our data (in the form of burnt): ranking 14th with smell + adjective (3rd subcorpus). The adjectives investigated in Pettersson-Traba’s study do not appear with smell at all.

The adjectives complementing stink are a mixed bunch. Though they mostly mark a smell as negative (bad, fierce), there is also a hit with the filler good. Generally, most of the adjectives amount to nothing more than expressing the evaluation of a smell, so that the receiver of the message does not get a clue with respect to the concrete quality of the smell at issue. The only exception is the adjective sour. On a more general level, Winter (cf. 2019: 237) investigated English sensory adjectives and observed similar tendencies in the usage of adjectives denoting smell. He points to a similar finding already made by Majid and Burenhult (2014: 266) from a more general linguistic perspective. Hence, it does not come as a surprise that the collexemes calculated for the verb smell basically confirm this finding, though with the proviso that smell seems to attract more positive evaluations than found elsewhere (11 out of 28 collexemes in period 2, and 10 out of 30 collexemes in period 3). As for the higher prominence of negative evaluations in the olfactory sense, Winter (2016) collected information on potential reasons for this. For example, he mentions the fact that humans cannot avoid (the perception of) smell (which seems most obvious in case of a bad-smell experience), or that they cannot exert control over smell. Moreover, he points to the effect of preconceptions in Western culture that smell “is an ‘animalistic’ or ‘primitive’ sense (…)” (Winter 2016: 983). Therefore, the increase in positive evaluations in our data from the most recent subcorpus may be a first indication of potential changes in this respect.

Looking at the particular adjectives attracted by the two verbs, we notice that stink (third period data) collocates with very general evaluative adjectives (good, bad) and with adjectives denoting the intensity of the smell (badly, fierce). The adjectives attracted by smell are much more diversified. For the third period data, the general evaluative adjectives good and bad top the list of collexemes, followed by adjectives implying more emotionally loaded evaluations (such as delicious (rank 3) or awful (rank 5). The list also comprises a number of adjectives giving information on the quality of the smell experienced, such as fishy (rank 9) and burned (rank 14). The data of the second period are fewer, with more adjectives rendering positive evaluations (good rank 2, lovely rank 6) and fewer adjectives on lower ranks giving qualitative information on the smell (scorched rank 7, foul rank 15).

As apparent from the examples (of stink) just cited, the data also contain hits with an adverb instead of an adjective: 2nd period: smell disagreeably, sweetly (next to sweet), and 3rd period: sweetly. This may be a reflection of older uses, as is suggested by quite a number of the examples collected in the OED. The earliest quotations of the verb smell have an adverb in the postverbal slot, which speaks for the utterance’s structure to be a combination of an intransitive construction complemented by an adverbial construction (of manner). Stink is found in English from very early on, the quotation being from AElfric Ic stince swote with later examples having swete and fule. As these forms could be used as both adjective and adverb, the word class is neither explicit, nor definite. These illustrations show that there seems to have been indifference to the form of the postverbal component:^[32] we find adverbs and adjectives alike, with the latter use increasing over time. This trend is also reflected in our data, which, being picked from American usage, will also display the more general tendency in informal (spoken) American English to use adjectives instead of adverbs (such as real good for British English really good).

In the following section, we turn to a discussion of the changes in the patterns over time as they are observable from the data, and a selective discussion of some differences in usage showing up between the same patterns found for the four smell verbs in one period.

4.3.1 Comparisons of smell-verb constructions between verbs

Within the scope of this paper, it is impossible to discuss all the verbs in all the constructions for the ways in which they are special or different from one another. That is why we just present exemplary results from the analyses.^[33] From the constructions we investigated, we will focus here on

1. smell verb + noun (subcorpus 2)

2. smell verb of noun (subcorpus 3)

3. smell verb of/like/with noun (suborpora 1 and 3)

4. smell verb + adjective (subcorpus 2)

In doing so, we will mainly concentrate on the constructions’ fillers themselves and relate them to the domains, as done for the collexeme interpretation.

1. smell verb + N (1920–1929)

The data from this subcorpus contain uses of only 3 of the four smell verbs investigated, with reek attracting only 1 filler (plan), which makes the use metaphorical. The remaining hits pertain to the verbs smell and scent.

The verb scent attracts in the top ranks such distinctive fillers as story, mystery, danger, plot, making this construction prone to metaphorical uses (16 out of 20 fillers), which communicate that a particular situation or phenomenon is felt to be approaching. This finding confirms what was already indicated by the verb’s attracted collexemes. The filler air (domain ‘world’) in the top rank points to the fact that the double role of the filler noun in this pattern can barely be overlooked: it may be the stimulus or the descriptum in a smell perception (cf Section 4.2.2). The remaining fillers name substances/objects (automobile, bouquet) and living things (deer). As for evaluations of the smells people scent, the construction seems to prefer value-free nouns in the literal uses. However, the metaphorical uses almost exclusively refer to disliked states or phenomena.

The verb smell is special in that it attracts pronominal fillers (it, them, him, what), suggesting that the thing smelled is not a first mention in the communicative context (except for the last example). The remaining fillers are a mixed bunch, words denoting substances (smoke, gas, powder) or relating to the body (perfume, breath, hair) are more prominent than words naming living things (rose, blossom) or food or agricultural items (bacon, hay). It is noticeable that, not unlike the collexemes, most of the 20 top distinctive fillers do not appear to add evaluative information on the kind of smell perceived. This, however, may well be a skewed result of our analysis, since the fillers (as said in Section 3.4) were investigated neglecting the modifiers they may have co-occured with in the utterance. The same reservations apply to what was said on the fillers of scent.

1. smell verb of N (2010–2019)

The ‘smell-verb + of N’ construction is a construction in which all four smell verbs occur in the 3rd subcorpus, though for scent we could extract only 3 hits. Table 6 provides an overview.

Starting with ‘reek of N’, the top distinctive collexemes (sweat, smoke, beef stew, money) exhibit a variety of domains. These are reinforced by the fillers to follow, which relate to the body (body odor, burned flesh), refer to food (ale, chardonnay), and substances (charcoal, cigarette smoke), or trigger a metaphorical reading (arrogance, defeat, despair). As for fillers of the last type, we can observe an increase here, so that, in the most recent data, the ‘reek of N’ pattern seems to take over the lead in metaphorical usage from the ‘scent + N’ pattern, which was more prominent in the 2nd period. Looking at the smell stimuli employed as fillers in literal instantiations, we find that they tend to express strong smells, with some of them even being socially stigmatized (such as sweat, burned flesh, cigarette smoke).

‘Scent of N’ is rare, the few collexemes turning up as distinctive all refer to living things/subtances (rosemary, flower, cedar), naming plants that emit intensive scents. In our culture, these are perceived as very strong and pleasant and they play an important role in the supply of fragrances and flavors to the cosmetics industries.

‘smell of N’ stands out by attracting quite a few substance words (soap, gasoline, disinfectant, balsam, paint) as distinctive, next to body related expressions (cologne, aftershave) and words naming food products (coffee, milk). There are no metaphorical readings induced by the top distinctive fillers. As compared with ‘reek of N’, the evaluation of the smells causing the smell experience tends to be much more positive (cologne, soap, aftershave, earth), with only a few fillers naming smells that are experienced as bad (but not socially stigmatized: gasoline, disinfectant, mothballs or paint).

Finally, in ‘stink of N’, the sources of the smells are mainly negatively perceived (fish, animal waste, body odor), some clearly belong into the group of taboos (urine, shit). There are also metaphorical “sources” of smells (condescension, illiteracy, laziness), all of them stigmatized in our culture.

Summing up the above, we can say that in the smell-verb of N construction reek and stink are mainly concerned with negative smell experiences, the other two verbs (smell and scent) are more hospitable to the description of positively perceived or evaluated smells.

1. smell verb with/of/like N (1820-1829)

This comparison aims at 3 formally different, but semantically similar constructions, which differ in the smell verb used. For the first period investigated here, the number of hits is limited, but still the analysis brought to light some preferences in the individual patterns. Since in the subcorpus we only found hits for scent with, reek with, smell like and smell of, we just look at these 4 different constructions. The results of the computation were quite informative about the peculiarities of each (Table 7).

Firstly, reek with is noticeably special with the filler blood.^[34] It also attracts fever and smoke, marking it as mainly expressing negative smells. Secondly, scent with is connected with positively evaluated smells, such as blossoming wilderness. Also vinegar, probably associated with the function of one of the few cleaners or disinfectants available at that time, can be assumed to have had a positive evaluation. Thirdly, smell like attracts hay as filler noun (positive evaluation), and, finally, smell of has stronger (and more negative) smell sources as fillers (oil, tar, onions). There is also one metaphorical use with mortality as filler.

1. smell verb + Adj (1920–1929)

This construction, the predicative construction with an adjectival complement, does not occur (in this subcorpus) with the verb reek, so that we look at only three smell verbs (Table 8).

In this period, the verb scent shows most clearly the attraction of both adjectives and adverbs (natural, awfully, nicely) (cf. our comments in Sections 3 and 4.2.4). The evaluations given by the adjectival fillers are mainly positive or rather pleasant (great, nicely, natural) but not exclusively so: some give unpleasant evaluations (awfully) or mark the smell as suspicious (curious).

‘smell + adjective’ is often connected with positive evaluations, qualifying the smells as pleasant (good, sweet/ly, lovely, delicious etc.), but unpleasant smells turn up as well. Their description is, however, less drastic (musty, dank, disagreeably). The few examples of stink in the pattern render negative smells (bad/ly, vilely). This distribution is in line with the findings made for the ‘stink like N’ and ‘reek with N’ constructions (cf. Section 4.2.3) and can be seen as an expression of the verbs’ negative semantic prosodies, or may even have contributed to their emergence.

4.3.2 Comparisons of individual smell-verb constructions across time periods

For the detection of changes in the uses of individual smell verbs over time, a multiple distinctive collexeme analysis can reveal potential differences in the fillers of the three time periods considered. Again, what follows here is an exemplary discussion of the results, also considering findings from the (simple) collexemes discussed in Sections 4.2.1–4.2.4.

We will look at

1. Smell + N

2. Smell of N

3. Smell like N

4. Smell + Adj

Generally, the lists computed by the programme are clearly different from the output of the (simple) collexeme analysis. In fact, some of the distinctive fillers exhibit noticeable differences in the ranks they have in the (simple) collexeme lists. To give an example, in pattern iv) the collexeme weird ranks 8th in the list of distinctive collexemes (3rd period) and 23rd in the respective collexeme list.

With this in mind, we will turn to the individual patterns.

1. smell + N over time

The top 20 collexemes in each period are provided in Table 9.

For the earliest period, this construction was found with 16 filler types only, so that all filler types (and the respective domains they represent) turn up in both types of analysis. The differences in ranking suggest that the pattern, in view of its later uses, distinguishes itself by the filler fragrance (rank 1 – (simple) collexeme rank 8). Moroever, it is used for the rendering of unspecified types of smell (elements, the one (ranks 2 and 4 vs. 12 and 13 in the list of simple collexemes)), where the specification must be assumed to be read off the context of the utterance. That means the smell descriptor cannot have been a new (rhematic) element in the respective discourse. Finally, it is also worth mentioning that the filler blood ranks 3rd in the first period, as against 14th in the (simple) collexeme list. This filler was already identified as prominent in this period, though as a collocate in the reek with pattern.

The distinctive fillers from the 2nd and 3rd periods follow the trend to refer to smell sources that are contextually given/known: (8) them, (13) us (2nd period), (1) her, (4) that, (7) me, (8) it, (15) you (3rd period). Except for that, none occurred among the top 30 collexemes of the respective periods.

Users in period 2 preferentially speak of smells from the domains of ‘substances’ (water, powder, gas), ‘food’ (liquor, hay, grub) and ’life’ (rose, blossom, bear).

Speakers from the 3rd period mainly employ such fillers as blood, scent, air, rose, fear, cologne (ranks 2, 3, 5, 6, 8, 10), relating to ‘things’ from various domains, with none being particularly prominent. In other words, speakers draw on their smell experiences in quite a number of different fields. This can be understood to mean that smell descriptions drawing on smell seem to become more variable and reflect changes in the smells people are exposed to. Apart from what was said on the role of the cosmetics industry above (cf. Sections 4.2.1 and 4.3.1), further such changes become apparent, for example, in the occurrence of the smell emitters marijuana and weed among the top 30 fillers distinctive for the 3rd period. These fillers hint at social changes in the more recent past regarding (the acceptance of) the consumption of new intoxicating substances.

As for smell evaluations that might go with the fillers used, the top 30 filler words (again – neglecting their potential modifiers) indicate a decrease in the number of those referring explicitly to a smell generally considered bad. At the same time, they comprise more offensive words (compare sulphur (period 1), grub (period 2), and shit, urine (period 3)).

In all 3 periods, the pattern does not stand out for metaphorical uses (one in the top 30 distinctive collexemes each) thief (1st period), news (2nd) and fear (3rd).

1. smell of N over time

This construction, with its 7 filler types in the earliest period, does not give much room for variation in their ranking. Compared with the simple collexeme list, camphor is a less prominent distinctive filler. Mortality, in contrast, gains in distinguishing power, marking this construction for metaphorical uses.

Table 10 provides a survey.

The first 3 fillers (on ranks 1 and 2) from period 2 are identical with the top 3 simple collexemes, but further substances (dust, ammonia, chloride) are distinctive for it, interspersed by food items (tea, bacon) and supplemented by living things (fish, herring, horses) further down in the list. The smell emitters identified include chemicals, perhaps as a reflection of their growing presence in people’s lives. As for the smell evaluations carried by the fillers, some name unpleasant smells, others are additionally rated negative because they are dangerous to people’s health (brimstone, soot, ammonia, chloride, gas). Only few fillers refer to smells that are generally perceived as pleasant (tea, Christmas, cleanliness (all of them ranking quite differently in the simple collexeme list (3:12, 4:44, 4:21)), and one makes the utterance metaphorical (inquisition).

Positively evaluated distinctive fillers are more frequent in the 3rd period (cologne, cinnamon, coffee (ranks 3 and 4), aftershave, rose-water, balsam (ranks 5 and 6)), all associated with the domains of body and substance. Fillers with negative associations also rank highly (cigarettes (3), gasoline (4)), taboo words appear on the lower ranks (piss (6), cat piss (8)). Hence, taking a diachronic perspective, we can observe two trends. The construction is used for the expression of an increasing number of pleasant smells, but at the same time it attracts more words marked as taboo. It remains stable in its rare attraction of metaphorical fillers.

1. smell like N over time

The top 20 distinctive collexemes for each period are given in Table 11.

In the first subcorpus, the construction appears just once, with a positively evaluated filler (hay).

In period 2, the distinctive fillers mainly come from the domains of living things (animal, herring, porgy (boat), food (corn, doughnut, ham) and substances (incense, isthagaria, sugar).^[35] All these fillers also appear in the simple collexeme lists, the ranking being quite similar (except that animal ranks lower (27)). Metaphorical readings are triggered by the fillers custom and rake.

In period 3, emitters of unpleasant smells sit high on the list (shit, smoke, blood, dead bodies, sweat, cigarettes and fart),^[36] and they pervade the whole list. That means that speakers nowadays appear to prefer smell like N for the descriptions of bad smells, and that, using quite offensive language, they feel less inhibited to use taboo words. The distinctive fillers naming pleasant smells refer to substances people regularly experience in modern days (perfume, coffee, lemon, vanilla, cinnamon). From this we can cautiously assume that, perhaps because these fillers are rather specific, they are quite effective smell descriptors and make sure that the quality of the smell is correctly communicated and recognized. Fillers triggering metaphorical readings are absent from this period.

1. smell + Adj over time

Table 12 provides a survey of the top 20 distinctive fillers.

The data of period 1 contain 2 hits only, so we focus on the 2nd and 3rd periods. In the second period, the top 3 distinctive fillers are positive (sweet, lovely, fine), with more of them following on lower ranks (pleasant, delightful). In period 3, however, the unpleasant smell emitters are more prominent (bad, awful, burned, fishy, rancid, disgusting).^[37] Some of them just express a speaker’s evaluation of a smell, but three of them additionally render qualitative information: burned, fishy (if used literally) and rancid. Such adjectives were also prominent in the 2nd period (7 out of 23 fillers vs. 6 out of 30 fillers in the 3rd period). As for metaphorical uses, the fillers alone are not sufficient to decide on this: for example, fishy (period 3) and foul or rotten (period 2) are open to both literal and metaphorical interpretations, depending on what kind of thing functions as the carrier of the smell (or stimulus). As this is usually expressed by the subjects of the utterances, this is a problem that can be solved by a (still outstanding) covarying collexeme analysis.

For reasons of the scope the paper, we will leave the discussion at that and draw some conclusions in the following section.