Nonsense Words Elicit Meaningful Drawings

Davis, Charles P., Hannah M. Morrow, and Gary Lupyan. 2019. “What Does a Horgous Look Like? Nonsense Words Elicit Meaningful Drawings.” PsyArXiv. April 2. doi:10.31234/osf.io/uve7d

Abstract: To what extent do people attribute meanings to “nonsense” words? How pervasive is such attribution? We used a set of seemingly meaningless words to elicit drawings of made-up creatures. Separate groups of participants rated the nonsense words and the drawings on several semantic dimensions, and were asked to choose a name for the drawn creatures. Nonsense words elicited a high level of consistency in the produced drawings. Meaning attributions made to nonsense words corresponded with meaning attributions made by separate people to drawings that were inspired by the name. Naïve participants were able to recover the name that inspired the drawing with greater-than-chance accuracy. These results suggest that people make liberal and consistent use of non-arbitrary relationships between forms and meanings.

---
1. Introduction
Imagine a group of artists illustrating children’s books about made-up creatures with
names like “horgous” and “keex.” Will different artists create similar drawings for creatures that
have similar names? Will readers who encounter the drawing of a “horgous” expect this creature
to be named “horgous”? We show here that drawings elicited by certain “nonsense” words relate
in a systematic way to the form of these words. This relationship between nonce words and the
meanings they express is bidirectional: Certain word-forms lead people to infuse their drawings
with certain properties. Other people, looking at the drawings, match them back to the original
names at higher than chance levels.
The idea that certain words fit some meanings better than others has its roots in the
ancient world (Plato, 1999) but was all but excised by the ascendance of structural linguistics (de
Saussure, 1959) and its focus on the sharp boundary between the signifier and the signified. The
conventional wisdom has been that with the exception of words that directly imitate sounds, the
relationship between word-forms and meanings is arbitrary: “There is no reason for you to call a
dog ‘dog’ rather than ‘cat’ except for the fact that everyone else is doing it” (Pinker & Bloom,
1990). In the last several decades, iconicity—a resemblance between form and meaning—is
being recognized as one of the basic design features of natural language in both the signed and
spoken modalities (Perniss & Vigliocco, 2014, Dingemanse et al., 2015, Monaghan et al., 2014,
cf. Hockett, 1978). The idea that the auditory modality can convey meanings in an iconic way,
beyond simple imitation of sounds, is at first counterintuitive. For example, Hockett argued that
the relationships between spoken words and meanings is arbitrary because “When a
representation of some four-dimensional hunk of life has to be compressed into the single
dimension of speech, most iconicity is necessarily squeezed out” (Hockett, 1978 p. 274). We now know that speech is a richly multi-dimensional signal, and spoken languages make ample
use of this dimensionality to convey meanings in an iconic way. For example, consonant voicing
(/b/ vs. /p/, /d/ vs. /t/) is used to signal differences in mass: Siwu: tsratsra, ‘a light person walking
quickly’ vs. dzradzra, ‘a heavy person walking quickly’. Vowel quality is used to signal size:
Ewe: lɛgɛɛ : logoo, ‘slim : fat’. Vowel lengthening is used to signal duration and intensity:
Japanese:  piQ : piiQ, ‘tear short : long strip of cloth. And reduplication is used to signal
repetition: Tamil: curuk-nu : curukcuruk-nu, ‘a sharp prick : many sharp pricks’ (Dingemanse et
al., 2015; see also Perniss, Thompson, & Vigliocco, 2010). Although these specific form-to
meaning relationships are not found in all languages, examining statistical relationships between
forms and meanings across languages does reveal some more universal relationships (e.g., Blasi
et al., 2016).
What makes such examples especially interesting is that people appear to be extremely
sensitive to form–meaning relationships even when the specific relationship is not present in
their language. For example, monolingual English 3-year-olds are sensitive to Japanese form–
meaning relationships when learning novel words (Kantartzis, Imai, & Kita, 2011). In English,
there does not appear to be a systematic relationship between object size and vowel pitch. Yet
anyone who has read to a small child knows that there is something exceedingly natural about
saying “elephant” in a lower pitch than “mouse,” or using a higher pitch to refer to the baby
elephant compared to the mommy elephant.1 Similarly, in English, smaller objects do not as a
rule have shorter names than larger objects (indeed, the relative shortness of “whale” compared
to “micro-organism” is used by Hockett to illustrate the principle of arbitrariness; Hockett,
1978). And yet, when asked to select a nonce-word for a small object such as a pin, people not
only prefer shorter words, but justify their choices with statements like “a small item’s name
should be small” and “pins are sharp and simple, as is this word” (Lupyan & Casasanto, 2012).
Hearing adults with no sign language experience can also use iconicity to determine the
approximate meaning of a word (i.e., a word’s telic content—whether it refers to an event with a
finite end point or not) across a number of signed languages, even when the exact meaning of the
sign is not a response option. People are more likely to choose “believe” than “forget” when
presented with a sign for “think” (Strickland et al., 2015), perhaps because telic signs make event
boundaries salient, while atelic signs are characterized by repetition and a lack of such
boundaries.
In an early systematic investigation of what he called “phonetic symbolism,” Sapir
(1929) recruited over 500 participants of varying ages (mostly, but not exclusively, English
speakers) and presented them with dozens of short nonce words and asked to distinguish them on
size. For example, a participant may be told that “mal” and “mil” both mean table; they then had
to decide which would be a better word for a large table. The chosen answer, overwhelmingly
and largely independent of age and language background, was “mal.”2 Sapir concluded that most
people displayed a common “feeling-significance” toward vocalic and consonantal contrasts in
nonsense words and that it made “surprisingly little difference whether the phonetic contrast was
contained in a [phonotactically] ‘possible’ [i.e., attested in the subject’s language] or …
‘impossible’ context” (p. 228). Sapir did not know where these sound-to-meaning mappings
came from, but speculated that they may arise from people implicitly learning that producing
 certain vowels requires larger mouth cavities. This early speculation was greatly amplified by
Ramachandran and Hubbard’s (2001) replication of Sapir’s phonetic symbolism demonstration
(see also Kohler, 1929; Newman, 1931) giving us the well-known “bouba-kiki” effect wherein
people overwhelmingly match “bouba” to a round shape and “kiki” to an angular one (see also
e.g., Maurer, Pathman, & Mondloch, 2006).
Contemporary investigations of sound symbolism have not settled the question of where
these associations between forms and meanings come from (but see e.g., Imai & Kita, 2014;
Sidhu & Pexman, 2018; Spence, 2011), but they have further demonstrated the varied way in
which iconicity plays a role in language learning and vocal communication. For example, Perry
et al. (2018) showed that more iconic words are learned earlier by children (adjusting for
numerous potential confounds like frequency, concreteness, and communicative need; see also
e.g., Imai, Kita, Nagumo, & Okada, 2008; Maurer et al., 2006; Peña, Mehler, & Nespor, 2011;
for further review, see Imai & Kita, 2014). Such apparent advantages of iconicity go beyond
word-learning. For example, people think that someone named Bob ought to have a round face
while a Mike should have a more angular one (Sidhu & Pexman, 2015). Further, Lupyan and
Casasanto (2015) had people learn to categorize two kinds of “aliens.” One of these contained
aliens that were subtly more round and the other subtly more pointy. When the categories were
labeled with the nonce words “foove” (which people tend to associate with being round and
friendly) and “crelch” (pointy and dangerous), people learned the category distinction itself (not
just the category names) better than when arbitrary or iconically incongruent labels were used.
When tasked with creating novel vocalizations to communicate a range of meanings (e.g., big,
small, high, low, smooth, rough, cook, fire, fruit, and many others), people not only converge on
surprisingly similar vocal forms, but when these vocalizations are played to naïve listeners
(including those from other language backgrounds), they are understood at levels well above
chance (Perlman & Lupyan, 2018).