Rethinking body partonymy: transverse analogous paradigms across world languages

1 Introduction

Finger bone connected to the toe bone,

Foot bone connected to the hand bone,

Wrist bone connected to the ankle bone

Leg bone connected to the arm bone…

Relationships like these might seem doubly dubious, first as basic misconstruals of anatomy and physiology, and second as the corruption of an old folksong. But what if such relations actually hold true conceptually across world languages? That is the possibility under consideration in this paper, focused on a comparative corpus of linguistic evidence drawn from a representative sample of 50 world languages to recover what many have assumed to be a lost cause in body partonym studies: empirical grounds for shared conceptualization cross-linguistically. To accomplish this, the study shifts attention away from isolated body parts connected visually or kinesthetically in linear or hierarchical arrays to focus instead on transverse embodied mappings: analogous correspondences and patterns that hold between upper and lower body partonymy. The overall blending paradigm (and paradigmatic blend) that emerges may actually reflect an early and enduring model of human creative cognition, as I discuss further below. For now, to aid discussion, I refer to this model as “transverse analogous paradigms,” or TAP for short. As I clarify further in what follows, evidence for TAP may include anything from fingers and toes (as separate English lexemes for analogous transverse concepts) to nails, digits and limbs (as identical English lexemes for transverse analogous concepts), and much more.

My guiding hypothesis is that most world languages, regardless of family, feature overt evidence of TAP at the lexical level, not only in isolated matches but also in paradigm sets. If so, this would mean that tapping into TAP across world languages stands to bring us closer to discovering the shared human understanding of perceived relationships between bodily parts and wholes (and part-whole relations in general) than any merely linear, fragmented or hierarchical model could account for. Exchanging the current linear comparative model (think: foot bone connected to the leg bone) for a non-linear analogic approach (think: foot bone connected to the hand bone) requires a fundamental rethinking of comparative body partonymy itself, not only in terms of experience and cognition (from phenomenology to psychology), but also in terms of our study of embodied partonymy as an experiential and cognitive phenomenon (from theory to methodology). Notably, a sea change in comparative body partonymy is long overdue.

The search for cross-linguistic correspondence patterns in comparative body partonym studies, has been underway since the questing heydays of the 1970s launched searches for linguistic universals in this domain of perception and cognition (Brown, 1976; Andersen, 1978; Brown and Witkowski, 1981). Brown's (1976) pioneering study not only launched “partonymy” as the namesake rubric of this pursuit but also cast the cross-linguistic search for correspondences in this domain as an enterprise focused on hierarchical patterns of part-whole relations. In this approach to partonymy, a superordinate term like “body” was expected to be composed of nested parts like “head” and “limbs,” each of which was expected to be composed of its own respective nested parts in turn: heads being composed of “eyes,” “nose,” “mouth” and so on; eyes being composed of “lids,” “eyelashes,” “eyeballs,” and so on. But problems quickly surfaced. Some languages have no word for head or even body. Others seem to pay meager attention to individual body parts at all. Still others have extremely detailed lexicons of body parts and their nested memberships (Majid, 2015), and mismatches between languages resist even the most basic typological mappings.

Speakers of Kewa from Papua New Guinea (Trans-New Guinea, Franklin, 1963), much like speakers of Hlepho from China (Tibeto-Burman, Pelkey, 2011a,b), for example, both use a single lexeme (ʦ^*h*Black handwritten numeral one with a curved top stroke on a white background.²¹ and aa, respectively) to refer to the lower limb which English speakers separate into “leg” and “foot.” The pair of terms “hand” and “arm” in English also contrast with composite lexemes in Kewa and Hlepho (lε³³pe⁵⁵ and kíi, respectively), such that what an English speaker would take to be one's “hand” is extended to cover the entire length of one's “arm” in the other two languages. While Kewa features a colexification for “sole” and “palm” (wáraa), English and Hlepho distinguish the two using separate partonyms. In contrast with both English and Hlepho, Kewa speakers use colexifications for many other body parts that are otherwise distinct, including “ankle joint” and “elbow joint” (ká-kálo), “heel” and “elbow” (kínyálú), “toe” and “finger” (kilikili), “toenail” and “fingernail” (kídípaa). And while English and Phowa speakers also use colexifications to denote the hard coverings on both toe-tips and finger-tips (nails in English, se^ in Hlepho), the Hlepho term se^ is polysemous with “claw” (applying to both humans and alloanimals), while the English term is not. Furthermore, while Phowa uses ʦ^*h*Black handwritten numeral one with a curved top stroke on a white background.²¹ for both “wrist” and “ankle,” English uses different terms for each; and there is no “ankle” term in Kewa, only kéréop “wrist.” Such variation starts to seem random, especially when scaled up to a more representative sample of world languages.

Due to problems like these, patterns claimed early-on as cross-linguistic universals in this area came to be undermined by the results of empirical research on the lexical semantics of body parts around the world. Asifa Majid and colleague's landmark special issue of Language Sciences (Majid et al., 2006; Enfield et al., 2006) was especially decisive in this assessment. Comparing the studies compiled in the collection revealed such widespread lexical variation in this domain, and so little evidence for shared hierarchies or conceptual models at the lexical semantic level, that the whole enterprise soon came to seem questionable. In Enfield's own assessment, the widespread lack of correspondence even casts doubt on the possibility that human beings might have any shared conceptual system of part-whole relationships at all (Enfield, 2006, p. 199). This stance poses a distinct challenge to theories of embodied cognition.

In the years that followed, study after study added evidence for the sheer diversity of strategies for the lexical identification and organization of the bodily semantic field around the world and their resistance to neat comparative correspondences across languages (Brenzinger and Kraska-Szlenk, 2014; Majid, 2015; Kraska-Szlenk, 2020; Tjuka et al., 2024). Soon it became apparent that this was true even within the same language family (Huisman et al., 2021) and even between speakers of the same language (Majid and van Staden, 2015; Devylder et al., 2020).

What seems to have been missed in the critique of proposed semantic universals, lexical hierarchies and cross-linguistic conceptions in body partonymy, is an alternative perspective oriented to wholes in relation, or rather parallels between part-whole sets. The missing perspective in question has had little uptake, but it is not new or absent altogether. In fact, it was also put forward during the 1970s, thanks to Roy Ellen's observation that “there is generally a recognition of similarity between the upper and lower parts of the body, reflected in identifiable structural and terminological correspondences between upper and lower limbs” (Ellen, 1977, p. 349). In Ellen's assessment, the field's neglect of this shared conceptual source stems from a kind of theoretical and methodological bias: “there has been a tendency to stress the function of individual parts rather than the relationships between them, either anatomically or functionally” (1977, p. 364). Remarkably, the same holds true to this day. Reorienting the search for lexical evidence of shared bodily templates requires a shift in attention from isolated parts to paradigmatic relations. As I have argued elsewhere, the implications are potentially profound for both theory building and analysis, with apparent implications for theorizing the origins of language and creative imagination and for understanding the nature of human cognition and ideology (Pelkey, 2016, 2017, 2018, 2022).

Beyond noticing parallels between arms and legs, as Ellen recommends, the shift in attention that this requires, a shift that the present study presupposes, is a fundamental return to one of the grounds of human evolution: the emergence of upright posture and its concomitant reorientation of our anatomical planes and directional axes relative to all other animals (Alexander, 2004). As Van Lier notes (2003; 2010), standing and walking habitually at an orthogonal angle relative to the earth resulted in the emergence of a transversalizing primate, with upper-lower relations (such as upper and lower limbs) thrown into bold relief, potentially enabling the emergence of transverse analogous paradigms across the waistline and their affordances for creative analogical relations in general. I return to these possibilities in Section 4. First it will be most helpful to turn to the study itself.

2 Materials and methods

In order to begin testing for the relative degrees of TAP across world languages, a representative sample of world languages is needed, along with a working database of body part terms coded according to operational definitions that are useful and valid for identifying lexically transparent dimensions of conceptual contrast and comparison.

2.1 Sampling, screening, and database compilation

To guard against the skewing effects of typological diffusion, methodological cherry picking, and confirmation bias, the study attempts to sample a diverse array of languages at a size still manageable for manual coding, featuring languages of both greater and lesser diffusion. To these ends, the study relies on published lexicographic data from 50 world languages, compiled with the help of two funded research assistants (see Sections Acknowledgments and Funding). Candidate languages are quota sampled according to three dimensions of contrast: geographic distribution, genetic linguistic diversity, and speaker population. After compiling a preliminary list of languages according to these criteria, research design shifted to quality screening to ensure that only languages with detailed lexicographic sources would be included in the comparative database. Candidate languages with only simple wordlists or generic translations available as lexical data were omitted and replaced with a viable alternate. Candidate languages with lexicographic sources featuring definitions, notes, examples, and/or clarifying illustrations were considered ideal. Data collection included all body-part terms, along with notes on morpheme glosses, translations, and variant forms. See Table 1 for lexicographic sources.

Table 1. Language sampling criteria and distributions.

In spite of these screening measures, some degree of variation in terms of quality and reliability between sources is accepted as inevitable. Anthropological linguistic studies focused on body partonymy in a given language, such as Franklin (1963) and Gaby (2006), can be considered the most thorough and reliable for a comparative research project like this. Research-oriented dictionaries such as Evans (2019) and François (2021) are also ideally suited. Ordinary bilingual dictionaries still tend to provide a wealth of evidence, though; and this is true even of the two oldest sources considered in the project, both of which are published in the late nineteenth century (Larajasse, 1897; Roberts, 1895). While languages change, the level of detail needed for valid analysis in this project is still relatively general. These degrees of variation are accepted as a limitation of the study due to constraints and requirements of sampling, feasibility, and data availability.

The resulting database includes 50 languages from at least 46 major sub-branches of 34 distinct language families, spoken on five continents and five macro-regions around the world. General locations of languages sampled are illustrated in the Figure 1 map, with numbers corresponding to the detailed listing of information on languages sampled in Tables 1, 2. See Supplementary material combined tables.

Table 2. Language data sources and coding compilation.

^*Piro(Yine) is supplemented by Bartlett and Hodge (1909); Mi'kmaq is supplemented by Rand (1988).

The sample includes 15 languages of greater diffusion, 15 languages of moderate diffusion, and 20 languages of lesser diffusion, with the third category primarily including 14 languages of minimal diffusion. Minimal diffusion is defined as having first-language (L1) speaker populations ranging from 50 to 10,000 L1 speakers. Languages of moderate diffusion are defined as ranging from one million to fifty million L1 speakers, and languages of greater diffusion range from 60 million to nearly one billion L1 speakers.

2.2 Coding the database

Database coding proceeded in four steps: (1) data sorting, (2) initial analysis, (3) intercoder reliability testing, and (4) revised analysis. First, in order to focus on upper-lower limb relations, the study sorted body part entries for each language to feature upper limb (UL) and lower limb (LL) comparisons. Any lower-limb lexeme (from hip to toe anatomically) was included in LL, while any upper limb lexeme (from shoulder to finger anatomically) was included in UL. In the next step, I applied three layers of coding simultaneously and iteratively, as described below. For the third step, intercoder reliability testing, I relied on the support of a funded research assistant (see Sections Acknowledgments and Funding). The dialogue that resulted led to the definition of a fourth layer of coding, transverse limb matching, as described below. First consider the three primary levels of coding:

• Level 1. Transverse parallel conceptions: Analogous matches above and below the waistline in spite of mismatched lexemes.
• Level 2. Transverse colexifications: Analogous matches using identical lexemes above and below the waistline.
• Level 3. Transverse paradigm sets: Two or more instances of Level 2 mapping onto one or more instances of Level 1 to form a diagrammatic paradigm set (with patterned, internal iconicity between part-whole relationships) above and below the waistline.

Coding for Levels 1 and 2 is more qualitative while Level 3 coding is assigned automatically when multiple instances of Level 2 are apparent, but all three levels function together as an integrated implicational hierarchy of evidence for part-whole membership organization since Level 3 necessarily contains instances of 1 and 2, while 2 is simultaneously an instance of 1.

Level 1 coding identifies transverse parallel conceptions. This level of TAP is suggested by the presence of separate lexemes for transverse analogous limb relations, such as “elbow” and “knee,” “ankle” and “wrist,” “fingers” and “toes.” While this level of coding is the most common among languages sampled, and perhaps the least reliable as conclusive evidence, it should not be overlooked, since it implies the same kind of attention to transverse analogous relations and differentiations as more advanced levels. In other words, the difference is one of degree, not kind.

Level 2 coding identifies instances of transverse colexification. For purposes of this study, “colexification” judgments are restricted to language-internal domain comparisons across the transverse plane only. As such, English nail and Phowa se^ are both considered transverse colexifications since they each apply to limb relations above and below the transverse plane. Kewa kínyálú is also a colexification, in this case of “elbow” and “heel” (one above and one below the transverse plane), regardless of the fact that this relation is unprecedented elsewhere.

Hlepho ʦ^*h*Black handwritten numeral one with a curved top stroke on a white background.²¹ “foot/leg,” by contrast, is not considered to be an instance of colexification because Hlepho speakers identify the corresponding limb relation below the waist using a different lexeme: lε³³pe⁵⁵ “hand/arm.” This brings up an important distinction in contrast to typical applications of “colexification” in the literature that would lead many to consider both terms to “colexify” concepts that English speakers keep lexically separate (see e.g., Rzymski et al., 2020; Tjuka et al., 2024; Tjuka, 2025). Whether or not it may be warranted elsewhere, such an analysis would be both invalid and irrelevant here, not only because it would involve assumptions of the very kind of natural classes and universal hierarchies that have been the Achilles heel of the enterprise for the past 50 years, but also because it would distract from this study's focus on language-internal parallels between part-whole relations. As such, colexifications are only considered to exist language-internally in this study and only relative to a clear plane of contrast: in this case the transverse anatomical plane.

Even so, it is also important to note that Hlepho ʦ^*h*Black handwritten numeral one with a curved top stroke on a white background.²¹ “foot/leg” and lε³³pe⁵⁵ “hand/arm” match conceptually, in terms of transverse analogical relations, in spite of their lexical difference. The same can be said for Kewa aa “foot/leg” and kíi “hand/arm.” English “foot” vs. “leg” and “hand” vs. arm” by comparison also match across the transverse plane, but in a clearly different sense: i.e., by both being separated. These relations call for a fourth level of coding: Transverse Limb Matching (LM). Although this level of coding is in one sense a sub-set of the first coding category, the concepts in question are more salient than other examples in this category since they are by definition the most basic or fundamental for comparative transverse limb relations. As such, they are also coded separately for purposes of this study, with three possible distinctions: Matching Composite (MC), Matching Separate (MS), and Non-Matching (NM). See the LMCode in Table 2.

Both Kewa and Phowa are coded as featuring MC (matching composite) limb relations in the database, as described above, in contrast to Matching Separate (MS) limb relations in languages like English and Zulu that separate “hand” and “arm” above to match the separation of “foot” and “leg” below. Languages that show no match in limb relations at this level are coded as non-matching (NM). An example of this is notable in Georgian, which features composite relations above and separate relations below: compare pekhi ფეხი “leg/foot” vs. kheli ხელი “hand” and mk'lavi მკლავი “arm.”

To illustrate this coding process in action, consider the abbreviated dataset from Teop (Austronesian, Papua New Guinea) based on Ulrike (2019):

Table 3.

Each of the first three pairs ([moo, kuri]; [takatakaku mooman, takatakaku kurinaa]; [pasuka moo, isuku kuri]) are coded as Level 1: transverse parallel conceptions. Level 2 applies to kobisi “finger/toe” as a transverse colexification. Since isuku “joint” applies to both elbows and knees according to Ulrike (2019), it potentially qualifies for Level 2 coding as well, and takatakaku is also analyzed as part of a transverse colexification construction, thus qualifying it as a hybrid of Level 1 and Level 2. As such we have abundant evidence even in this small dataset for Level 3 transverse paradigmatic relations: i.e., two or more instances of Level 2 mapping onto one or more instances of Level 1 to form a diagrammatic paradigm set due to sharing patterned, internal iconicity between part-whole relationships above and below the waistline.

As a fourth layer of coding for Teop, overall limb relations are classified as Matching Composite (MC) since the lower limb foot/leg composite moo matches the upper limb hand/arm composite kuri. In summary then, Teop is coded as MC-3 (matching composite limb relations with three nested levels of TAP). Added together these features qualify the language with four degrees of evidence for transverse analogous relations. Following these same procedures, coding results for each language are listed in the Table 2 summary.

The Degree column in Table 2 is simply a combination of LMCode- and TAPCode- values. Considering levels of TAP as degrees of evidence, the presence of limb matching in a given language (whether MC or MS) provides an additional degree of evidence. The absence of limb matching in a given language (NM), by contrast, slightly decreases the overall degree of evidence or confidence. As such, languages like Navajo and Choctaw that feature three levels of TAP in addition to limb matching can be rated with an even higher degree evidence for TAP relations than languages like Guaraní and Somali, which feature three levels of TAP without limb matching. These relationships are clarified further in the next section, along with further rationale for the related distinctions.

3 Results

Coding for these four dimensions of contrast in the 50 languages sampled provides diagnostic leverage for testing the hypothesis presented above: i.e., that most world languages, regardless of family, feature overt evidence of transverse analogous relations at the lexical level in paradigm sets. As a preliminary observation for purposes of analysis, it is important to clarify that these four levels or dimensions of contrast cannot be considered equally diagnostic as sources of evidence for TAP. Level 1 (transverse parallel conception) constitutes weak evidence for TAP; Level 2 (transverse colexification) constitutes moderate evidence; and Level 3 (transverse paradigm sets) constitutes strong evidence. Level 3 is also the most complex, as discussed above. The three levels or dimensions in relation suggest an integrated implicational hierarchy of evidence for part-whole membership organization.

Dimension 4 (transverse limb matching) by comparison is additive instead of implicational. As such, I define the presence of transverse limb matching in a given language (whether MC or MS) as confidence raising in four potential degrees: e.g., by raising the overall TAP evidence of a Level-1 coded language to the second degree (from 25% confidence to 50% confidence), raising a Level-2 coded language to the third degree (from 50% confidence to 75% confidence), and raising the TAP evidence of a Level-3 coded language to the fourth degree (from 75% confidence to 100% confidence). With these operational definitions and relations in mind, it is possible to survey the findings of database coding and to analyze results.

3.1 Implicational hierarchy findings: Levels 1–3

Starting first with evidence for TAP relations from the three levels implicational hierarchy, all languages in the database show at least moderate evidence of TAP through transverse parallel conceptions (Level 1) and transverse colexifications (Level 2). Only three languages in the database (6%) stop there. The remaining languages (n = 47, 94%) also show strong evidence of TAP through transverse paradigm sets, as defined above.

The three languages with only Level-1 and Level-2 evidence in the database are French, Finnish, and Mi‘kmaq, each from widely separate backgrounds in terms of the four layers of sampling criteria presented in Section 2 above. In each case only one instance of transverse colexification is apparent in the data. In French and Mi'kmaq this happens to be the transverse colexification for “nail” (French ongle, Mi‘kmaq ugqosi'l), applying to both toes and fingers. Finnish is more of a borderline case, potentially qualifying as Level 3. In the Finnish data, kämmen “palm (of hand) / forefoot” is a clear instance of Level 2 transverse colexification, and a novel lexicalized metaphor pair serves as further evidence that nonetheless falls outside the operational criteria as a second instance of Level 2. Finnish jalkaterä “foot” is a compound construction of “leg+blade,” while käsivarsi “arm” is a compound construction of “hand+stem/shaft.” While this does not count as colexification proper, it provides further evidence of transverse relations at the level of analogous lexicalization, drawing on metaphorical object mappings (see Tjuka et al., 2024 for further discussion) which would arguably qualify it for paradigmatic relations in a broader sense. In fact many other related sources of additional evidence can be cited across languages for TAP, as I discuss briefly in Section 4.

3.2 Level-4 confidence-ratings: transverse limb matching

First, it is helpful to factor in the additional evidence of limb matching (NM, MS, MC), considered as an additive or subtractive degree measure of TAP relations. As described above, the presence or absence of limb matching serves as supplementary evidence for interpreting levels of TAP supplied by implicational hierarchy relations. MS and MC values Added to these three possible degrees, limb matching values serve as an additive confidence rating, potentially boosting implicational cline values (for MS and MC languages) or slightly downgrading overall confidence (for NM languages) accordingly. Applying limb matching confidence ratings to the languages sampled, all but one (n = 49, 98%) show strong evidence for TAP relations with at least 75% confidence rating (three out of four degrees), and 76% (n = 38) of languages sampled show comprehensive indices of TAP (four out of four degrees or 100% confidence rating). The remaining outlier, Mi'kmaq shows a modest confidence rating of 50% (two out of four degrees).

Accordingly, most languages sampled feature analogous lexical matches between transverse limbs, with 40 out of 50 languages (80%) rated as matching (MS or MC) and 10 out of 50 (20%) rated as non-matching (NM). Within the 80% of languages that feature matching limb relations, it is worth noting that 60% (n = 24) qualify as MC (composite transverse matches of [“hand/arm”] + [“foot/leg”]) and 40% (n = 16) qualify as MS (separate transverse matches of [“hand” + “arm”] + [“foot” + “leg”]). This evidence suggests that composite transverse matches are more common among world languages than their separated counterparts, while also supporting any suggestion that Anglo-centric conceptions of limb relations are not the norm.

Of the three languages in the database without adequate evidence of Level-3 TAP relations, only Mi'kmaq is coded as NM, resulting in a 50% confidence rating overall. French and Finnish are boosted from 50% to 75% confidence since both feature transverse limb matching lexically. Finnish qualifies as an MC language; while French qualifies as an MS language. The nine other languages without clear evidence of transverse limb matching (NM) still show strong evidence of TAP elsewhere, especially through Level-2 transverse colexifications. The Kuuk Thaayorre language of Australia (Pama-Nyungan), for example, features distinct terms for foot and leg parts, and these match with parallel arm and hand parts; but arm parts are clearly lexicalized using a derivative construction of punth, which is a general term for “arm,” while there is no such general term for “leg” (Gaby, 2006, p. 212). Nonetheless, Kuuk Thayore shows robust evidence for Level-3 TAP relations, with at least four clear instances of Level-2 transverse colexifications, as summarized in Table 3.

Table 3. Transverse colexifications in Kuuk Thayore (Gaby, 2006).

Similarly in Tibetan, a composite lexicalization of “foot/leg” is apparent below the waist, but “arm” vs. “hand” are distinct lexicalizations above the waist (compare rkań-pa “foot/leg,” with phyag “hand,” dpung-pa “arm”). Even so, no less than six instances of transverse colexification emerge between upper and lower limbs in Tibetan, including thuń-wa “lower-arm/leg,” ńar “forepart.of.lower.leg/arm,” mthe-bon “pinkie.finger/little.toe,” sen-mo “nail (of finger/toe),” the-bo “thumb” ~ mthe-bo “big.toe,” and more. Other NM languages show similar patterns. Somali and Japanese in turn are the most ambiguous for LM coding but also have the most transverse colexifications in the current database, with seven instances each.

With these examples in mind, it is worth considering strategies for ranking a given language higher in terms of TAP relation quality if it features more than two instances of transverse colexification. That possibility lies beyond the scope of this paper for reasons described in the next section; for purposes of analysis here, suffice it to say that limb matching is an inadequate measure of TAP on its own. This fact provides further justification for using it as a supplementary metric of degree instead as I have done here. In other words, since many of NM-coded languages show more than two instances of Level-2 TAP relations, Level-4 is an inadequate measure on its own and can be treated instead as an additive or subtractive confidence metric.

4 Discussion

All world languages sampled show some evidence of TAP, and the vast majority (94%, n = 47) show strong evidence through Level-3 transverse paradigm sets. This in itself is conclusive evidence in support of the hypothesis of the study, and even the three partial exceptions are instructive, especially with the addition of Level-4 confidence ratings involving transverse limb matching, which increases evidence for TAP even further. Considering transverse limb-matching as an additional measure, 80% of languages surveyed show further evidence of TAP at this level, notably including two of the three languages that did not qualify for Level-3 TAP relations. Treating Level-4 as a degree-of-confidence metric results in strong evidence (three out of four degrees) for TAP relations in 98% (n = 49) of world languages sampled, and comprehensive evidence (four out of four degrees) for TAP relations in 76% (n = 38) of world languages sampled. By all measures considered, then, the hypothesis tested in this study is confirmed: most world languages, regardless of family, feature overt evidence of TAP at the lexical level, not only in isolated matches but also in paradigm sets.In light of these results, it now seems fair to predict that all world languages show some degree of evidence for TAP, making it at least a universal tendency among world languages, if not a true universal.

4.1 Theoretical and methodological implications

This suggests that Ellen's (1977) early critique of body partonym research was both correct and ahead of its time. Looking for lexical evidence of cross-linguistic conceptual congruence at the level of individual parts and externally imputed hierarchies has mainly led to dead ends, for embodiment theory in particular. The field should refocus instead on transverse relations that hold between body partonyms and their implicational hierarchies relative to language-internal relations. These results also suggest that researchers who have assumed that the findings of body partonymy research undermine strong claims of embodied cognition might be working from false premises and invalid conclusions. Far from casting doubt on the possibility that human beings have any shared conceptual system of part-whole relationships at all (a la Enfield, 2006, p. 199), rethinking body partonymy in terms of transverse analogous paradigms reveals the presence of a general or universal bodily template of analogous affordances that might serve as a key source of the human ability to generate novel analogies.

As I have begun to argue elsewhere (e.g., Pelkey, 2017, 2023), the reorganization of the anatomical planes and axes of orientation intrinsic to human upright posture provides a plausible source of embodied evolutionary grounding for the emergence of advanced blending, argued by some to be the ability underlying the human language faculty (Fauconnier and Turner, 2002, 2008; Turner, 2014). The origins of this ability are under-theorized, especially from an embodied cognitive perspective. While the origins and motivations of bipedal upright posture are likely to be numerous and complex, ranging from threat display strategies and sexual selection to food gathering advantages, tool-use facility, and more (DeSilva, 2021; Hewes, 1961; Hunt, 1996; Lovejoy, 1981; Pelkey, 2017; Wheeler, 1991; Wescott, 1967), it is also worth considering the relationship between the evolution of habitual upright posture as a physiological, functional, survival-oriented phenomenon and the affordances or consequences of this evolutionary watershed for the emergence of human cognition.

While fore-limb/hind-limb specialization is common among other animals as well, the default caudal-rostral or front-back orientation parallel to the earth determines that the default function of both limbs is forward mobility. This also means that both limbs come in regular, repeated contact with the earth. Reorienting bodily planes of contrast and axes of orientation relative to the earth meant that hands not only became more specialized but also ceased being used for the same purpose as legs by default. Legs also became much more visible from a first-person experiential perspective. The possibilities for thinking reflexively and reversibly of arms in terms of legs and legs in terms of arms then would have become more available or possible. At the same time it would have been clearer than ever due to their dramatically marked differences that the two sets of limbs are nonetheless curiously similar, with nested layers of relations. The fact that all world languages seem to preserve observations of such relations across the transverse plane in analogous sets with paradigmatic iconicity both conceptually and lexically is a reasonable source of evidence for just such an emergence of awareness. Such affordances may then have seamlessly provided a natural, experiential basis for more advanced blending activities to build on. In other words, the fact that world languages show universal tendencies for systematic transverse analogous mapping between arm and leg relations may prove to be a key source of evidence for developing conceptual blending theory in dialogue with embodied cognition.

Since the point of advanced blending is not to distinguish between source and target mappings, in contrast with conceptual metaphors, or single-scope blends, but to map select conceptual possibilities between two mental spaces simultaneously, TAP relations also provide plausible sources for the formation of paradigmatic modeling more generally; and they also make room for both radical variation and universal constraints required for linguistic modeling. This would help develop the cognitive foundations of grammar put forward by Heine (1997), who finds that the bodily sources of cognitive grammar privilege the upper body for human understanding, while the lower body contributes only by being understood in terms of the upper (think of counting with fingers vs. counting with toes for instance). I have called these assumptions into question elsewhere (Pelkey, 2017, 2018), partly by drawing attention to clear counterexamples across world languages; and the findings of the current study call them into question further.

4.2 Limitations and future directions

Much work remains, naturally. This is only a beginning. The languages considered here are a representative sample, but the sample is small. In addition to expanding the study of arm-leg relations and testing the TAP hypothesis further, other partonym systems should be treated relative to frames of embodied contrast. More attention should be given to transverse colexifications, both in terms of possible typologies and in terms of qualitative measurement relative to other sources of evidence in addition to those considered in this study.

Other sources of evidence for TAP include intrafield and interfield meronymic mappings, interfield meronymic paradigms, ambidirectional historical intrafield shifts, and unified historical intrafield shifts, many of which are treated in summary fashion by Wilkins (1996), who discusses the shifts in question as “attested semantic changes involving visible parts and visible wholes” (1996, p. 276), without registering transverse analogical congruences in the data (see also Pelkey, 2017, p. 162–170).

Reorienting the field to the experiential nature of partonym paradigms is also a priority. Beyond embodiment in a superficial or abstracted sense, the diagrammatic iconicity of partonym paradigms is necessarily experiential or aesthetic in the extended sense of the term (Pelkey, 2022), helping explain how a model of embodied relations could become a modeling capacity or ability when extended or mapped onto other domains. Given the imprecise and playful nature of creative analogy, this modeling capacity would be expected to lead to widely divergent outcomes. This in turn would help show how the experiential basis of embodied cognition informs patterned relations between body partonymy and embodied mapping cross-linguistically. While the present study may offer supporting evidence for broader cognitive dynamics and evolutionary origins, it does not test these directly.

5 Conclusion

Even if the old song gets it wrong from a cross-linguistic perspective, linear models focused on isolated concepts, or isolated domains, are currently the status quo in body partonym studies; and attempts to identify cross-linguistic patterns in comparative body part studies have so far generated more noise than signal. The evidence presented in this study suggests that a reorientation to the comparative experiential relations of human upright posture can help the field tune in again to signals of cross-linguistic conceptual congruence. This congruence is not to be sought in comparisons between individual parts or externally imputed hierarchies but in the study of transverse analogous paradigms.

Applied to arm-leg relationships across a representative sample of world languages, the model that emerges provides further evidence of the cross-linguistic nature of embodied cognition through which human bodily movement, body memory, and sensory-kinesthetic feeling provide layers of shared experiential grounding that enable communication and translation in spite of differences across languages and cultures. The universal frame, pattern, or template in question here is the lived human body, engaged in self-reflection by noticing paradigmatic analogies between arms and legs despite their morphological, anatomical and functional differences. As such, rethinking body partonymy in terms of transverse analogous paradigms may also provide theoretical grounding for the emergence and evolution of language through creative analogy and grammar-oriented abilities.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/flang.2026.1937860/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Omid Khatin-Zadeh, University of Electronic Science and Technology of China, China

Reviewed by: Zahra Eskandari, International University of Chabahar, Iran

Jie Gao, Kyoto University, Japan