Skip to navigation – Site map

HomeIssues2PapersSubmorphemic iconicity in the lex...


Submorphemic iconicity in the lexicon: a diachronic approach to English ‘gn- words’

Dennis Philps


By applying a principle of submorphemic invariance to the word-initial segment gn- (phonologically /n/) in the heuristic class of English ‘gn- words’, I show that this segment functions as a ‘sublexical marker’ (<gn->), defined as a semiologically and notionally invariant submorphemic unit within the subset(s) of words of which it is a formative. I go on to claim, based on evidence provided by phononotional GN-/GVN(-) alternation, that although <gn-> must be considered as structurally indivisible in Modern English, <g> in <gn-> may originally have functioned as a ‘core invariant’ and <n> as a variable in the pre-history of English, specifically in Proto-Indo-European. Analysing gn- in most ‘gn- words’ of Anglo-Saxon origin and some of Greek origin as the surface trace of an underlying notional domain linked to jaw-related phenomena (sound articulation, mastication, manducation, etc.) suggests that there exists an extralinguistically motivated, causal relation between <gn-> and this domain involving the concept of ‘submorphemic iconicity’.

Top of page

Full text

I would like to express my grateful thanks to two anonymous readers for their comments on an earlier version of this paper.

1. Introduction: from phonæstheme to sublexical marker

1Phonæsthemes (Firth 1930) are frequently recurring, systematic pairings of form and meaning in the lexicons of many languages, including English. Their morphological status is probably best characterized as ‘submorphemic’, if one accepts Crystal’s definition of a submorpheme as ‘a part of a morpheme that has recurrent form and meaning, such as the sl- beginning of slimy, slug, etc.’ (Crystal 2003: 301). It may be argued that phonæsthemes have submorphemic status because they do not combine with morphemes to form words, but with submorphemes (e.g. sn- + -iff > sniff), while morphemes either form words themselves, or combine with other morphemes to do so by means of morphological processes such as inflection, derivation, and compounding. However, this status is the subject of ongoing debate (see Bloomfield 1933: 245, Bolinger 1950: 130, Bergen 2004: 290, etc.). The application of principles of invariance to phonæsthemes has led me to reanalyse them as ‘sublexical markers’, defined as semiologically and notionally invariant submorphemic units within certain subsets of the word classes in which they occur (‘sn- words’, etc., see Philps 2002), and noted between chevrons (<sn->, etc.).

2In this paper, I shall start by setting up a heuristic class I call ‘gn- words’, on the basis of a structural analysis of the GNVC(-) lexical schema assumed to underlie the stems of these words. This class comprises all and only those lexical items that begin with gn-, phonologically /n/ in Modern English. I shall then look at phononotional CR-/CVR(-) alternation, e.g. br-/bVr(-) in bring / bear (common notion: ‘carrying’), before going on to examine one particular sublexical marker, the analysis of which poses the problem of the link between semiological invariance and phonological invariance, namely <gn->. I use the term ‘phononotional’ to denote frequently recurring, systematic pairings of form and notion at submorphemic level.

  • 1 The Oxford Dictionary of English, 2nd edition, 1989.

3Few scholars have attributed phonæsthemic status to gn- in English ‘gn- words’, possibly because it is no longer pronounced as a cluster in Modern English, along with the fact that studies relating to phonæsthemes often have a synchronic bias. However, it is empirically verifiable that most ‘gn- words’ of Germanic origin and some etymologically unrelated ‘gn- words’ of Greek ancestry have meanings referring to jaw-related phenomena such as the manner of articulating sounds (e.g. dial. gnatter ‘to grumble, complain’ [grumble (v.) ‘to utter dull inarticulate sounds’, OED1], mastication (e.g. gnaw ‘to bite…’), and manducation (e.g. dial. gnip ‘to eat’), as shown in fig. 1, which will serve as our corpus.

Fig. 1: English ‘gn- words’ denoting jaw-related phenomena

  • 2 The English Dialect Dictionary (Wright 1898-1905).
  • 3 The New Shorter Oxford English Dictionary, 1993 edition.

gn- words’




‘to gnaw, bite hard at something, nibble, tear, bite’



‘to bite in a snapping fashion’



‘to snarl’



‘to growl, snarl, like an angry dog or beast; (fig.) to quarrel’



‘to strike together or ‘grind’ the teeth; to bite upon’



‘any of numerous small mostly biting dipteran insects…’



‘of or pertaining to the jaws’



‘to gnaw, bite at anything hard, nibble; to grumble, complain’



‘to bite (something) persistently…’



‘to take little bites of (a thing)…’



‘to eat; to complain constantly about’


4The fact that some ‘gn- words’ do not denote jaw-related phenomena may be explained in part by what Bolinger (1949) calls ‘cross-influence’, i.e. analogically driven blending, as in the case of gn- and sn- words from Proto-Germanic onwards, e.g. gnarl ‘to contort, twist…’ < *g(e)n- ‘to compress into a ball’ (Watkins 2000: 26) / snarl ‘tangle, twist together confusedly’ < *(s)ner- ‘to wind, twist’ (Pokorny 1959: 975), and in part to the fact that they derive from PIE roots which do not convey jaw-related notions (e.g. gnosis < Gk. gnôsis < PIE *gneh3- ‘to know’).

5Finally, I shall attempt to determine whether <gn->, like <sn-> (Philps 2002), contains a core invariant, and will invoke the concept of ‘submorphemic iconicity’ (see Bottineau 2003 for that of ‘semantic iconicity’) to account for the invariant notional domain apparently vectored by gn- in the subset of ‘gn- words’ in question.

2. Phononotional CR-/CVR(-) alternation

6If words beginning with gn- are set up heuristically as a ‘class’, and if derivational phenomena are set aside (e.g. gnag > gnaggle ‘to gnaw, bite at anything hard’, EDDII), then the members of this small class may be represented schematically as having an underlying GNVC(-) structure (where V = any possible vowel, C = any possible consonant, and (-) = suffixable), since all ‘gn- words’ have stems that conform to this schema. In the various realizations of GNVC(-), -V- can be long, e.g. in dial. gneep ‘a foolish fellow’ (EDDII), short, e.g. in gnat, or a diphthong, e.g. in gneiss; if -V- is long, -C can be Ø, e.g. in Standard English gnaw.

  • 4 Gnap (v.) ‘to gnaw, bite at, nibble, snap at’; gnip (v.) ‘to eat’ (EDDII).

7It may then be posited that any formal difference attested by the ‘rime’ (Bolinger 1950) results from a transformation operating on the variable segment of the GNVC(-) lexical schema (-VC) when realized, while GN- remains invariant insofar as it is the defining condition of the class. The segment -VC is analysed as a variable segment in GNVC(-) on the basis that when gnap and gnip4 for instance, are opposed, the transformation in question modifies -V-; when gnat and gnap are opposed, it modifies -C, and finally, when gnat and gnip are opposed, it modifies both -V- and -C (see fig. 2 below).

Figure 2: Structural invariance and variability in ‘gn- words’

Figure 2: Structural invariance and variability in ‘gn- words’
  • 5 Where S = s and C = voiceless occlusive [an older word for ‘plosive’, Trask (1996)], or resonant.
  • 6 Where, simplifying a little, C = voiced occlusive, voiceless occlusive, voiceless fricative, or ø ( (...)

8In word-initial position, there are two major types of consonant segments considered to be phonæsthemes in English: SC- segments (sl-, sn-, sp-, sw-, etc.),5 and CR- segments (br-, fl-, gl-, tr-, etc.).6 Now I have shown elsewhere (Philps 2002) that in certain subsets of the phonæstheme sn-, C- functions as a ‘core invariant’, defined as the smallest structural unit within a given subset of a given class of words to which a common notion may be attributed, and s as an initial variant. This analysis is based on the fact that the presence ([+s]) or absence ([-s]) of s- in phonosemantic ‘sn-/n- doublets’ such as sneeze / neeze (dial.) or sniff / niff (fam.) does not modify the basic meaning, as may be seen in fig. 3 below.

Figure 3: Dictionary definitions of phonosemantic ‘sn-/n- doublets’


sn- doublets’

Principal meaning of [+s] form

Principal meaning of [-s] form


snag (v2.): ‘to snap, bite’


nag (v.) ‘to gnaw, nibble’ (OED);

‘to bite’ (EDDIV)


snar (v.) ‘of a dog: to snarl, growl’


narr (v.) ‘of dogs, etc.: to snarl or

growl’ (OED)


sneeze (v). ‘to drive or emit air or breath

suddenly through the nose and mouth…’


neeze (v. dial.) = sneeze (v.)



sniff (v.) ‘to draw air through the nose

with short or sharp audible inhalations’


niff (v.) ‘to have a disagreeable

smell’ (OED)


snob (n.1) ‘one who wishes to be regarded

as a person of social importance’ (OED)

nob (n.3) ‘a person of some social

distinction’ (OED)


snuzzle (v. dial.) ‘of a dog: to sniff or

poke with the nose’ (OED)

nuzzle (v.1) ‘to poke or push with

the nose’ (OED)

Sources: OED, EDD

9The above analysis is inspired by the French linguist Gustave Guillaume (1883-1960), who hypothesized that ‘the various contextual senses of a morpheme arise as actualized meanings from a single potential meaning’ (Hirtle, in Guillaume 1984: XII; see also Hirtle 2007: 21-22). Guillaume saw the causal distinction between condition and consequence as the universal principle governing the construction of all languages: “[T]he linguistic sign establishes an invariant condition in tongue [Fr. ‘langue’] from which there develops an infinite number of consequences” (1964: 150, my translation).

10Transposing these hypotheses to the submorphemic level allows one to show that invariant word-initial phonæsthemes such as sn- in ‘sn- words’ and, as I argue here, gn- in ‘gn- words’, are endowed with a potential for meaning. This potential can only be realized linguistically once the underlying notion vectored by the invariant submorpheme, e.g. jaw-related phenomena in one subset of ‘gn- words’, is particularized by submorphemic accretion to such a degree that it attains meaningful status, e.g. gn- + -awgnaw. CR- segments, on the other hand, require an approach which differs in one respect from that applied to SC- segments, in that they display phononotional CR-/CVR(-) alternation rather than phonosemantic SC-/C- alternation, as may be observed in fig. 4 below.

Fig. 4: Phononotional CR-/CVR(-) alternation

CR- segments

Doublets / Common notion

bl- (/bl/)

{blare / bawl (v./v.), shouting}

br- (/br/)

{bring / bear (v./v.1), carrying}

cl- (/kl/)

{clamour / call (v.1/v.), shouting}

cr- (/kr/)

{crêpe / curl (v./v.1), curling}

dr- (/dr/)

{drab / dark (adj./adj.), (absence of) colour}

fl- (/fl/)

{flame / fulgent (n./adj.), brillance}

fr- (/fr/)

{front / fore (n./n.), anteriority}

gl- (/gl/)

{glut / gullet (n.2/n.), swallowing}

pl- (/pl/)

{plane / pallet (n.3/n.3), flatness}

pr- (/pr/)

{prime / paramount (adj./adj.), superiority}

tr- (/tr/)

{tree / tar (n./n.1), wood}

Source: OED

11By envisaging sublexical markers as surface traces of general underlying notions, one is able to attribute a ‘notional’ value to these markers within the subsets in which they occur, e.g. {(s)n-, nasality} in that subset of ‘sn- words’ including snaffle, sneeze, sniff, snob, snore, etc., or {gr-, prehension} in that subset of ‘gr- words’ including grab, grapple, grip, grope, grub, etc. Within the present framework, a ‘notion’ is envisaged as a conceptual space which may be defined as a set of complex mental representations resulting from the mind’s attempts to categorize experience, notably the formal, functional and compositional properties of the latter. These attempts are necessarily conditioned by human subjectivity, culture and convention.

12Seen in this light, notions may be analysed following principled cognitive criteria into ‘notional domains’, noted between slashes (e.g. /nasality/), to which a metric may be applied (Culioli 1990: 69). Such domains are topologically constrained, insofar as they can be identified interactively in terms of an interior (I) and an exterior (E) separated by a boundary (B) which is flexible and permeable. Furthermore, they may be organized hierarchically on the basis of dimensional, meronymic, and other relational criteria. For instance, a notional domain may be a superordinate conceptual space (Σsup) including several subordinate spaces (Σsuba…n), while Σsup itself may be subordinated to a superordinate space (ΣSUP).

13The domain of particular interest to us here, that of jaw-related phenomena, has been set up heuristically on the basis that it is a salient conceptual space located within the superordinate domain of the human body, and encompasses several subordinate domains, e.g. the jaws, the chin [‘the front of the lower jaw’], the mouth and the teeth, as well as the topological, functional and expressive properties―both articulatory and gestural―associated with these subordinate domains. As Ladefoged recalls, movements of vocal organs, which he conceives of as gestures, are controlled and purposeful; in particular, the movements of jaw and tongue are interlinked:

The principal muscle of the tongue is the genioglossus, which pulls the back and root of the tongue towards the front of the mandible (the jawbone). The net effect of contracting the genioglossus muscle is that the tongue gets compressed within the jaw. (Ladefoged 2005: 125)

3. The sublexical marker <gn->

14While the orthographical forms of word-initial SC- segments always correspond to the word-initial phonological structure /SC/ in English (e.g. sn- = /sn/, sp- = /sp/, sw- = /sw/), this is not always the case with CR- segments, since although bl- corresponds to /bl/ and gr- to /gr/, for instance, the orthographical form gn- no longer corresponds to the word-initial phonological structure /gn/ as it did in earlier stages of the language (Dobson 1968: 977-978), but to /n/: the form gnaw, for example, is pronounced /nO:/ in late Modern English. Indeed, Jespersen has referred to initial gn- as a ‘consonant group’ which has become ‘lightened’ (1994: 4), while other linguists have examined the reduction of initial kn and gn to /n/, e.g. Kökeritz (1945), Dobson (1968), and Lutz (1992). A distinction must therefore be maintained between the synchronic and the diachronic status of these markers.

15As seen above, it is possible to bring into focus relations between certain members of the class of ‘gn- words’ and ‘gVn(-) words’ which would otherwise remain unobserved or appear fortuitous, by adopting a more abstract representation of these classes. This is accomplished by setting up underlying lexical schemas, e.g. GNVC(-) and GVN(-), remembering that different phonetic and phonological constraints (initial consonant reduction, vowel length, diphthonging, absence of a final consonant, variations due to rhoticity, etc.) apply to their respective realizations. One such relation is that between ‘gn- words’ and ‘gVn(-) words’ denoting jaw-related phenomena (mastication, manducation, manners of speaking, expressive gestures, etc.). Two schemas are involved here: GNVC(-), as in gnathic (adj.) ‘of or pertaining to the jaws’, and GVN(-), as in genial (adj.2 in OED) in the anatomical sense of ‘of or pertaining to the chin’. The ‘gn- words’ of Anglo-Saxon stock concerned, most of which are probably late analogical innovations, are often described as ‘onomatopœic’ and are of uncertain etymology. They may be assumed to continue Proto-Germanic *gn- roots’, the Indo-European forms of which are usually given as ‘hypothetical’, e.g. *gh(e)n- > Proto-Germ. *gnagan (> OE gnagan > Mod. Eng. gnaw, Watkins 2000: 29).

16It should be noted that <gn-> must, in certain circumstances, be analysed in conjunction with the sublexical marker <kn->, represented semiologically by kn-, which is also realized as /n/ in Late Modern English (e.g. knee: /ni:/). This is because certain English ‘kn- words’ continue *gn- clusters in PIE, e.g. knee < OE cnēo < Germanic *knewam < PIE *ĝneu-, base II of *ĝenu- ‘knee’. This root is also continued in English, via Latin, by genuflect, reflecting the lexical schema GVN(-), but additionally by knop ‘the rounded protuberance formed by the front of the knee or the elbow-joint’ and knuckle (< Germ. *knapp- and *knuk- respectively), both of which appear to derive from, or at least to be modelled on, the hypothetical *g(e)n-.

17As previously mentioned, the phonetic realization of G in the lexical schema GNVC(-) may not be instantiated in a given state of the language, due to the consequences of phonetic change. In ‘gn- words’ of Anglo-Saxon origin, for instance, the devoicing and eventual loss of the word-initial voiced velar occlusive appears to have begun during the Middle English period, judging by the dates of the earliest attestations provided by OED. For instance, gnat (< OE gnæt(t)), one of the rare ‘gn- words’ of Old and Middle English not to have undergone s- accretion (Philps 2003), was spelt knatt(e) in the course of the 14th century. This reduction affected educated speech as early as the 16th century, according to Dobson:

Initial [gn] had two developments which affected educated speech in the sixteenth and seventeenth centuries. In the first the [g] was lost by a direct process of assimilation to [n] […]. In the second, [gn] by dissimilation becomes [kn] […]; thereafter it develops with original kn- through [χn] and [hn] to [n*] and [n]. (1968: 977-8)

18To exemplify the first development ([gn] > [n]), Dobson quotes sources of the period (‘homophone’ pairings, etc.) which suggest that gnash had become nash, gnat > nat, and gnaw > naw, while the second development ([gn] > [kn] > [n]) is testified to by lists or pairings including gnat > knat and gnaw > knaw.

4. Core invariance and English <gn->

19One consequence of the phonotactically-driven reduction of /gn/ to /n/ for the diachronic analysis of <gn-> is that, contrary to <n> in <sn->, <n> in <gn-> cannot be analysed as a core invariant, since gnag, knag, and nag, for example, are diachronic evolutions of the same word, whereas sneeze and neeze, for instance, are different words (sneeze is first attested c1493 as replacing OE fnese, while neeze, from ON hnjósa, is attested earlier, c1340, indicating that it is not the result of consonant reduction affecting sn-). Hence phonosemantic gn-/kn-/øn- alternation cannot be equated with phonosemantic sn-/øn- alternation. Also, there is no ‘*g-mobile’ alternation in Indo-European comparable to the morphophonological ‘*s- mobile’ alternation on which the phonosemantic sn-/øn- ‘doublets’ in English (sniff / niff, etc.) and other Germanic languages are modelled (Southern 1999: 70-71). Phonosemantic ‘doublets’ such as gnar(r) (v., dial.) ‘to snarl, growl’ / narr (v., dial.) ‘of a dog: to snarl, growl’ must therefore have a different structural origin.

20In fig. 5 below, I have set up a class of ten lexical items (‘gVC(-) words’), based on the heuristic criteria that all correspond to the lexical schema GVC(-) subsuming GVN(-), and that all have meanings referring to jaw-related phenomena in either their nominal or their verbal uses, or both: parts of the anatomy (mouth, teeth, jaws, chin, gums…), expressive facial activities (opening the mouth, showing the teeth, gaping…), functional activities (biting, swallowing, breathing…), etc., in human and non-human animates. Within this class, three items correspond to the schema GVN(-): gan, gen and gen(-) in genial, while the others correspond respectively to GVB(-) (gob), GVL(-) (gill), GVM(-) (gam, gum), GVP(-) (gape), and GVR(-) (gur- in gurgitate). Given that the invariant segment in this class appears to be G-, since -V- and -C- are both variables, I will hypothesize that G- functions as a core invariant (<g>) here, as it corresponds to the smallest structural unit within the class of ‘gVC(-) words’ in fig. 5 to which the jaw-related notional domain common to all its members may be attributed.

21Support for this analysis is provided indirectly by the fact that substituting other possible consonants for g- in the class of ‘gVn(-) words’ (e.g. : p-: pen, t-: tan, l-: line, m-: man, etc.), does not allow one to identify a common notional domain other than fortuitously (e.g. dental (adj.) ‘of or pertaining to the teeth…’ (OED) and mental (adj.2) ‘pertaining to the chin’ (OED), but tan and pan, den and men, fin and sin, etc.). It can however be seen that much depends on the criteria used to establish this kind of class. For instance, if a criterion of semantic identity rather than notional identity were used, it is clear that the meaning of ‘mouth’, ‘the opening in the head of a person or animal through which food is ingested and vocal sounds emitted’ (gam) and ‘throat’, ‘the front of the neck beneath the chin and above the collar-bones’ (gan) would not allow these two items to be grouped together on the basis of semic commonality, unless evidence provided by polysemy were also taken into account (e.g. gam (n.) ‘large teeth, tusks, the mouth, the lips’), gan (n.) ‘the mouth, the throat, (pl.) the gums, the jaws’).

Fig. 5: English ‘gVC(-) words’ referring to jaw-related phenomena



Meaning(s) referring to

jaw-related phenomena


Meaning(s) referring to

jaw-related phenomena




1. pl. Large teeth, tusks

2. The mouth (OED)

The lips (EDDII)

To cause teeth to grow

crooked and overlapping

each other (EDDII)

? (cf. PIE *ĝembh- ‘tooth, nail’)


The mouth (OED)

The throat; pl. the

gums, also the jaws


To stare, gaze vacantly


(n.) cf. Sw. (dial.) gan, ‘the inner

part of the mouth, the mouth, the

gill of a fish’ (EDDII)

(v.) cf. Sw. (dial.) gana ‘to stand

and wonder at anything’ (EDDII),

cf. ON gan ‘yawn; cry’

<PIE *gheha-


The act of opening the

mouth; a yawn (OED)

To open the mouth wide,

esp. in order to bite or

swallow anything (OED)

(v.) ON gapa ‘to open the mouth’

< PIE *ĝheh2i- ‘to yawn, gape’




To open, gape; grin,

to show the teeth like

a dog (EDDII)

OE geonian (ginian) ‘to open the

mouth’ (EDDII) < PIE *ĝheh2i-

‘to yawn, gape’ (EIEC)

gen- in


(adj.2) Of or pertaining

to the chin (OED)


Gk. genus ‘jaw, chin’

< PIE *ĝenu- ‘jawbone, chin’



The lower jaw, the flesh

under the chin or ears;

the mouth, throat (EDDII);

the organ of respiration in

fishes… (OED)

1. To gut or clean (fish)

2. To handle the gills of,

take hold of by the gills


(n.) Middle Eng. gile, gill, from a

Scandinavian source akin to ON

*gil, gill of a fish < Germ. *geliz

< PIE *ghel-unā- ‘jaw’ (AHDIER)


The point of the jaw of a

fish (EDDII)

To open the mouth for

want of breath (EDDII);

to clean (fish) for curing


(n.) Compare Sw. (mun) gipa,

corner of the mouth (EDDII)

(v.) Origin unknown (OED)


The mouth (OED)

A beak (EDDII)

To spit (OED)

(n.) OF gobe, goube ‘mouthful,

lump’ < gober ‘swallow, gulp’

(? < Celt. *gobbo ‘mouth, beak’,

whence prob. Sc. and north.

Eng. gob mouth, ODEE)


The firm fleshy integument of the jaws and bases of the teeth (OED)

To deepen and enlarge

the spaces between the

teeth of (a worn saw)


OE gōma ‘palate, jaw’

< Germ. gō-ma

< PIE *ĝhēu- ‘to yawn, gape’


gur- in


(gurgitation) ‘swallowing;

guzzling’ (OED)

To swallow greedily or

immoderately (OED)

Lat. gurges ‘throat’ < PIE*gwrҰ-g-

< *gwerh3- ‘to swallow’ (AHDIER)

  • 7 The Oxford Dictionary of English Etymology (Onions 1966).
  • 8 The American Heritage Dictionary of Indo-European Roots (Watkins 2000).
  • 9 Encyclopedia of Indo-European Culture (Mallory & Adams 1997).


22It may be noted in passing that colloquial and dialect French also has a small number of etymologically unrelated ‘gn- words’ denoting jaw-related phenomena, in which gn- is pronounced /J/, e.g. gnaquer (v.) ‘to bite (into)’, gnaques (n. pl.) ‘the teeth (of a dog)’, gnagne (n.) ‘(large molar) tooth’, gnigue-gnaque (n.) ‘said of the movement of the teeth when chewing’, and gnac (n.) ‘bite’ (cf. metaphorical uses such as ‘avoir du gnac or ‘avoir la gnaque’), all late, onomatopœic formations. Modern French also has a few ‘gn- words’ whose meanings do not seem to be jaw-related, in which gn- is pronounced /gn/. These are either borrowings, e.g. from Italian, gnocchi (1864) or, from Hottentot, gnu (1778), or of Greek and/or Latin origin, e.g. gnome (1583), built to Gk. gnômê, and (also Lat.) gnomon (1547). This paucity is scarcely surprising since, as Ernout & Meillet state (2001: 278) with reference to gnārus, gn- was not preserved in Latin, while in Ancient Greek, apart from a few derivatives, one only finds γνάθος ‘jaw’ (< PIE *ĝenu- ‘jawbone, chin’) and γνάμπτω ‘to curve, bend’, of uncertain etymology (Chantraine 1999: 230). As for the lexical schema GVN(-), Modern French has ganache (< Ital. ganascia ‘jaw’) ‘±lower jaw (of a horse)’, gencive (n.) ‘gum’, and gingival (adj.) ‘relative to the gums’. While ganache appears to be related to Gk. γνάθος ‘jaw’ (but cf. γένυς ‘jaw’, also derived from PIE *ĝenu- ‘jawbone, chin’), gencive and gingival can be traced back to Lat. gĭnīva, whose form may display a trace of reduplication (*gen-g-īua) according to Ernout & Meillet (2001: 275).

5. Core invariance in Proto-Indo-European

23As seen above, data furnished by EDD and OED provide evidence of a notional link between the class of ‘gn- words’ denoting jaw-related phenomena such as mastication (e.g. gnaw), manducation (e.g. gnip), manners of speaking (e.g. gnap) and expressive gestures with bared teeth (e.g. gnarl) and that of ‘gVn(-) words’ denoting jaw-related phenomena. In the lexical schema underlying the ‘gn- words’ in fig. 1, i.e. GNVC(-), GN- has been identified as a core invariant, while in that underlying the ‘gVn(-) words’ in fig. 5, i.e. GVN(-), G- seems to function as a core invariant. Now if one accepts Benveniste’s reconstruction of the canonical form of the IE root as C1eC2 with two alternating thèmes or ‘bases’ (1935: 170), then it becomes apparent that English ‘gVn(-) words’ denoting jaw-related phenomena are modelled on base I (full grade) of the root (*CVC-), while ‘gn- words’ referring to the same notional domain are modelled on base II (zero grade, *CCV-). Also, the derivatives of one and the same PIE root form may be represented by both schemas in English, e.g. GVN(-) in genial ‘of or pertaining to the chin’, and GNVC(-) in gnathic ‘of or pertaining to the jaws’ (OED), both reflexes of *ĝenu- ‘jawbone, chin’, via Greek γένειον ‘chin’ and γνάθος ‘jaw’ respectively.

24Furthermore, according to one of the principles set out in Grimm’s Law, initial /g/ devoices to /k/ between Indo-European and Proto-Germanic (Szemerényi 1999: 17), while normally remaining voiced in Latin and Greek (Beekes 1995: 110). This helps to explain why the two lexical schemas corresponding to PIE *ĝenu- ‘knee’, namely GVN(-) (base I) and GNVC(-) (base II), are in complementary distribution in Eng. genuflexion (n.) and knee (n.): whereas the former continues, via late Latin, the GVN(-) base I schema, the latter continues, via Germanic, the GNVC(-) base II schema, becoming KNVC(-) in Late Modern English.

25Now there is structural evidence in PIE which suggests that in certain ‘root doublets’, the consonant occupying the C1 slot in the canonical root structure C1eC2 functions as a core invariant, and C2 as a variable. This evidence comes in two forms. The first of these is root-final *-r-/*-l- alternation (Benveniste 1964: 25) that does not correlate with a change in the underlying notion conveyed by the root, as in *gal-/*gar- ‘to call, shout’, *ĝhel-/*ĝher- ‘to shine’, *ghel-/*gher- ‘to call (out)’, *pel-/*per- ‘to strike’, *(s)kel-/*(s)ker- ‘to cut’, *wel-/*wer- ‘to turn’, etc., while the second is root-final voiced/voiceless alternation that, like *-r-/*-l- alternation, does not correlate with a change in the underlying notion, e.g. *peig-/*peik- ‘to cut, mark’, or *ĝebh-/*ĝeph- ‘jaw, ±eat’, etc.

26If the consonants occupying the C1 slot in the abovementioned PIE ‘root doublets’ with initial *ĝh-, *gh-, *ĝ- or *g- function as core invariants, and those occupying the C2 slot (*-r-/*-l-, *-g-/*-k-, *-bh-/*-ph-) as variables, then it is conceivable that the root-final consonants in other roots with initial *gh-, *ĝh-, *ĝ- or *g- whose derivatives denote jaw-related phenomena also function as variables, since they too enter into a relation of alternation which does not correlate with a change in the underlying notion conveyed by the root. Examples include *ĝembh- ‘tooth, nail’ (> Eng. comb), *ĝheh2i- ‘to yawn, gape’ (> Eng. gape), *ghel-unā- ‘jaw’ (> Eng. gill), the hypothetical roots *g(e)n- ‘to compress into a ball’ (> Eng. gnarl) and *gh(e)n- ‘to gnaw’ (> Eng. gnaw), *ĝenu- ‘jawbone, chin’ (> Eng. gnathic), and *gheha(u)-mr- ‘interior of mouth (e.g. gums, palate)’ (> OE gōma ‘inside of mouth or throat’, Mod. Eng. gum).

6. Conclusion: submorphemic iconicity in English ‘gn- words’

27To account for the dyadic relation {<gn->, /jaw-related phenomena/} that can be traced back, through Germanic or Hellenic, to Proto-Indo-European, two types of explanation are possible. One holds that since the linguistic sign has no natural connection with the referent, acquiring its value solely as a result of the differential, language-internal relations it maintains simultaneously with other, like signs in the system, it is senseless to identify possible links between the extralinguistic and the linguistic: as conceived by Saussure, the linguistic sign is, by definition, arbitrary. A second type of explanation has it that the sign has become arbitrary, implying that it enjoyed, at some emergent stage in the evolution of language, a partly natural relation with the extralinguistic. Of course, this stance has its problems, insofar as it is neither demonstrable nor falsifiable; furthermore, much depends on what is meant, exactly, by ‘natural’. Whatever the case, the hypothesis proposed below provides a possible account of this relation.

28The sublexical marker <g> assumed to function as a core invariant in the C1 slot of the above-mentioned PIE ‘root doublets’ and other roots with initial *gh-, *ĝh-, *ĝ-, or *g- whose derivatives have meanings referring to jaw-related phenomena such as sound articulation, mastication and manducation, is realized phonetically as a voiced velar occlusive, whether aspirated, e.g. *gh- in *gh(e)n- ‘to gnaw’, aspirated and palatalized, e.g. ĝh- in *ĝheh2i- ‘to yawn, gape’, palatalized, e.g. *ĝ- in *ĝenu- ‘jawbone, chin’, or not, e.g. *g- in *geh1i- ‘to sing’.

29By reanalysing this root-initial occlusive as a gesture of occlusion (Browman & Goldstein 1992: 156), and hypothesizing that it originally served, because of its occlusive quality, as an articulatory unit around which forms with meanings related to the goal-orientated, occlusive (open-close) mandibular movements involved in oral activities such as gnawing and primitive calling could be constructed, we may trace it back to a possible scenario of the emergence of human speech similar to that described by MacNeilage in his “Frame-Content Theory” (1998, cf. also MacNeilage 2008: Part V). MacNeilage sees the open-close mandibular cycle as the main articulatory ‘building block’ of speech production, and the evolution of the open-close alternation of the mouth for purposes of speech as “the tinkering of an already available motor cyclicity into use as a general purpose carrier wave for time-extended message production.” (1998: 506).

30In other words, I posit that the naturally occlusive properties of the various phonetic realizations of <g> originally endowed it with the capacity to refer back to the jaws, both metonymically―the (articulated) sound for the articulators―and iconically, in that the sign thus created partakes in the character of its object (Peirce 1991: 251). Evidence from Indo-European languages suggests that this putative self-referential capacity may have been extended, again by metonymical association (part/whole, interior/exterior, contiguity, etc.), to signs denoting the jaw’s physiological environment (chin, mouth, cheeks, teeth, gums, palate…), its movements (biting, gnawing, eating, swallowing…), and its expressive connotations (to gape [‘an open-mouthed stare’], to gaze [to stare (at)], to stare [‘to gaze fixedly and with eyes wide open]…), etc. The evolution of occlusions from the non linguistic to the linguistic, via the prelinguistic, may be modelled schematically as follows:

Fig. 6: Modelling the emergence and evolution of the sign

(Mandibular open-close cycles, notably for purposes of nutrition and non verbal communication)

(Sound gestures produced by mandibular open-close movements to stand iconically and metonymically for open-close actions linked to nutrition, such as gnawing, and those linked to facial expression of the emotions, such as gnashing the teeth, as well as for the articulators involved)

(Fully linguistic signs formed by syllabogenesis (C- > CV- / -VC) and subsequent morphogenesis (CV- > CVC / CVCV, etc.) for purposes of denoting, in speech, actions linked to nutrition and to facial expression of the emotions)

(Due in part to pragmatic/discourse factors such as usage and frequency, and in part to the drift of the underlying system towards the arbitrary (phonetic change, grammaticalisation, semantic change, etc.), the original, physical component of the meaning of the signs formed at stage C. begins to be ‘bleached out’, thus allowing these signs space to take on a new sense referring to, e.g., a manner of articulating speech rather than the act of articulating itself)

NB: ↓ = direction of evolution through time

31Hence, several millennia later, the meanings of English ‘gn-’ and ‘gVn(-)’ words referring to jaw-related phenomena (and also words having undergone affrication such as jowl, < OE cēafl ‘jaw, cheek’ < Germ. *kēfalaz < PIE *ĝebh-/*ĝeph- ‘jaw, ±eat’ and chin, < OE cin(n) < Germ. *kinnuz < PIE *ĝenu- ‘jawbone, chin’) would to some extent be predetermined by this causally based, self-referential capacity. However, this does not imply that such a possibility is necessarily materialized in any language, or that any realization of <g>, irrespective of its position, must actualize this potential. In Indo-European, for instance, *-ĝh- may function as a root extension (e.g. *bhelĝh- ‘to swell’ < *bhel-, Rix et al. 2001: 73-74), irrespective of the notion conveyed by the root in which it occurs.

32While reconciling these two types of explanation appears impossible, it may nevertheless be adduced that whereas English ‘gn- words’ and ‘gVn(-) words’ with jaw-related meanings are arbitrary at the morphemic level, they may be partially non-arbitrary at the submorphemic level. This would be so if the words in question were construed as being built to a core invariant (<g>) which originally possessed an inherent, self-referential value that gradually became more and more arbitrary as linguistic processes such as phonetic change (recall that /g/ in words beginning with gn- is no longer realized phonetically in Late Modern English) and structural evolution (gn- no longer appears to be functionally divisible in English) modified the original, semiogenetic conditions of the transformation of the sign from prelinguistic to linguistic status. If this postulate is admitted, then it may be concluded that a trace of submorphemic iconicity persists in English ‘gn-’ and ‘gVn(-)’ words denoting jaw-related phenomena.

Top of page


Beekes, Robert S. P., 1995. Comparative Indo-European Linguistics. An Introduction. Amsterdam / Philadelphia: John Benjamins.

Benveniste, Émile, 1964 [1939]. « Répartition des consonnes et phonologie du mot ». In Études phonologiques dédiées à la mémoire de M. le Prince N. S. Trubetzkoy. Alabama: University of Alabama Press, 27-35.

Bergen, Benjamin K., 2004. “The psychological reality of phonæsthemes”. Language 80 (2), 290-311.

Bloomfiled, Leonard, 1933. Language. London: Allen & Unwin.

Bolinger, Dwight L., 1949. “The sign is not arbitrary”. Boletín del Instituto Caro y Cuervo 5, 52-62.

Bolinger, Dwight L., 1950. “Rime, assonance and morpheme analysis”. Word 6 (2), 117-136.

Bottineau, Didier, 2003. « Iconicité, théorie du signe et typologie des langues ». In Le mot comme signe et comme image : lieux et enjeux de l’iconicité linguistique, Cahiers de linguistique analogique 1, Philippe Monneret (Ed.). Dijon: ABELL, 209-228.

Browman, Catherine P., & Goldstein, Louis M., 1992. “Articulatory phonology: an overview”. Phonetica 49, 155-180.

Chantraine, Pierre, 1999 [1968]. Dictionnaire étymologique de la langue grecque. Paris: Klincksieck.

Crystal, David, 2003. A Dictionary of Linguistics & Phonetics, 5th edition. Oxford: Blackwell.

Culioli, Antoine, 1990. Pour une linguistique de l’énonciation. Opérations et représentations, vol. 1. Gap: Ophrys.

Dobson, Eric J., 1968 [1957]. English Pronunciation 1500-1700, 2nd edition, 2 vols. Oxford: Clarendon.

Ernout, Alfred, & Meillet, Antoine, 2001 [1932]. Dictionnaire étymologique de la langue latine. Reprint of the revised 4th edition. Paris: Klincksieck.

Firth, John R., 1930. Speech. London: Ernest Benn.

Guillaume, Gustave, 1964. Langage et science du langage. Paris: A.-G. Nizet / Quebec: Les Presses de l’Université Laval.

Guillaume, Gustave, 1984. Foundations for a Science of Language (translated by W. Hirtle and J. Hewson). Current Issues in Linguistic Theory, vol. 31. Amsterdam: John Benjamins.

Hirtle, Walter, 2007. Language in the Mind. An Introduction to Guillaume’s Theory. Montreal: McGill-Queen’s University Press.

Jespersen, Otto, 1994 [1905]. Growth and Structure of the English Language, 10th edition. Oxford: Blackwell.

Kökeritz, Helge, 1945. “The reduction of initial kn and gn in English”. Language 21 (2), 77-86.

Ladefoged, Peter, 2005. Vowels and Consonants. 2nd edition. Oxford: Blackwell.

Lutz, Angelika, 1992. “Lexical and morphological consequences of phonotactic change in the history of English”. In History of Englishes: New Methods and Interpretations in Historical Linguistics, Matti Rissanen, Ossi Ihalainen, Terrtu Nevalainen, & Irma Taavitsainen (Eds.). Berlin / New York: Mouton de Gruyter, 156-166.

Macneilage, Peter F., 1998. “The frame/content theory of evolution of speech production”. Behavioral and Brain Sciences 21, 499-546.

Macneilage, Peter F., 2008. The Origin of Speech. Oxford: Oxford University Press.

Mallory, James P., & Adams, Douglas Q. (Eds.), 1997. Encyclopedia of Indo-European Culture. London and Chicago: Fitzroy Dearborn.

Onions, Charles T. (Ed.), 1966. The Oxford Dictionary of English Etymology. Oxford: Clarendon Press.

Peirce, Charles S., (1991 [1906]). “Prolegomena to an apology for pragmaticism”. In Peirce on Signs, J. Hoopes (Ed.). Chapel Hill: The University of North Carolina Press, 249-252.

Philps, Dennis, 2002. « Le concept de ‘marqueur sub-lexical’ et la notion d’invariant sémantique ». In La notion d’invariant sémantique, P. Larrivée (Ed.). Travaux de linguistique 45, 103-123.

Philps, Dennis, 2003. « S- incrémentiel et régénération submorphémique en anglais ». Bulletin de la Société de Linguistique de Paris XCVIII (1), 163-196.

Pokorny, Julius, 1959. Indogermanisches etymologisches Wörterbuch, 2 vols. Bern and Munich: Franke.

Rix, Helmut (Ed.), Kümmel, Martin, Zehnder, Thomas, Lipp, Reiner & Schirmer, Brigitte, 2001. Lexikon der indogermanischen Verben. Wiesbaden: Reichert.

Southern, Mark R. V., 1999. Sub-Grammatical Survival: Indo-European s- mobile and its Regeneration in Germanic. Washington DC: JIES 34 / Institute for the Study of Man.

Szemerényi, Oswald J. L., 1999. Introduction to Indo-European Linguistics. Oxford: Oxford University Press.

Trask, R. L., 1996. A Dictionary of Phonetics and Phonology. London: Routledge.

Watkins, Calvert, 2000. The American Heritage Dictionary of Indo-European Roots, 2nd edition. Boston: Houghton Mifflin.

Wright, Joseph (Ed.), 1898-1905. The English Dialect Dictionary, vols. I-VI. London: H. Frowde.

Top of page


1 The Oxford Dictionary of English, 2nd edition, 1989.

2 The English Dialect Dictionary (Wright 1898-1905).

3 The New Shorter Oxford English Dictionary, 1993 edition.

4 Gnap (v.) ‘to gnaw, bite at, nibble, snap at’; gnip (v.) ‘to eat’ (EDDII).

5 Where S = s and C = voiceless occlusive [an older word for ‘plosive’, Trask (1996)], or resonant.

6 Where, simplifying a little, C = voiced occlusive, voiceless occlusive, voiceless fricative, or ø (/g/ in /gn/ or /k/ in /kn/ in Modern English), and where R = resonant.

7 The Oxford Dictionary of English Etymology (Onions 1966).

8 The American Heritage Dictionary of Indo-European Roots (Watkins 2000).

9 Encyclopedia of Indo-European Culture (Mallory & Adams 1997).

Top of page

List of illustrations

Title Figure 2: Structural invariance and variability in ‘gn- words’
File image/png, 19k
Top of page


Electronic reference

Dennis Philps, Submorphemic iconicity in the lexicon: a diachronic approach to English ‘gn- words’Lexis [Online], 2 | 2008, Online since 10 November 2008, connection on 30 March 2023. URL:; DOI:

Top of page

About the author

Dennis Philps

Université de Toulouse-Le Mirail (France).

By this author

Top of page



Creative Commons - Attribution-NonCommercial-NoDerivatives 4.0 International - CC BY-NC-ND 4.0

Top of page
Search OpenEdition Search

You will be redirected to OpenEdition Search