How Does English Encode ‘Tight’ Vs. ‘Loose-fit’ Motion Events? It’s Complicated

ABSTRACT

Linguistic encoding of spatial events has long provided a forum for examining how languages encode space, how children learn their native encodings, and whether cross-linguistic differences affect non-linguistic representations of space. One prominent case concerns motion events in which objects are moved into tight or loose-fit relationships of containment or support. Seminal findings from Bowerman showed that young children learning Korean regularly use specific verbs to encode tight/loose fit across containment and support relationships, whereas children learning English use prepositions to encode containment or support (e.g. in/on) across the tight/loose fit distinction. Others have asked how these early-acquired differences affect non-linguistic encoding of similar events. Many of these studies have focused on the lexical differences between the two languages – verbs (in Korean) and/or prepositions (in English). Here, we ask whether this focus might underestimate how English encodes these events by closely examining the range of options used by English speakers to encode loose and tight-fit motion events. In Experiment 1, 3-year-old and adult English speakers described joining and separating events which culminated in loose or tight-fit end-states. Participants’ use of lexical verbs together with their syntactic frames differentiated among the event types, especially between “loose-fit” events with asymmetric motion between objects (e.g. a block being put into a bowl) vs. “tight-fit” events with symmetric motion (e.g. two Legos being brought together at the same time). In Experiment 2, we replicated the basic findings using events portrayed with more complex of objects. Our findings show that English affords both children and adults rich resources to encode motion events culminating in tight and loose fit end-states; these devices include both lexical items and syntactic frames. The findings raise important questions about how to examine effects of language on non-linguistic spatial cognition in children and adults.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Bowerman and Choi provided little data on the range of verbs used by English speakers for these motion events, but did indicate that put was used.

² In this paper, we refer to the words together and apart as symmetrical particles, as they logically entail two entities with a reciprocal spatial relationship: “She put X and Y together” entails that that X was put together with Y and vice versa. These words are unique in the class of spatial terms, which generally must be construed as asymmetrical; that is, spatial terms such as “in,” “on,” “above,” “behind” encode only asymmetrical relationships, e.g., if X is “above Y,” then Y cannot be “above X.” Symmetrical verbs are much more common in English; see L. Gleitman et al. (1996) for full characterization of symmetrical terms including verbs, spatial terms, and nouns.

³ This frame predominantly elicited a plural NP among adults (e.g., she connected the Legos) but sometimes did not, e.g., she assembled the egg; but in this and other cases, the singular NP implies multiple pieces which formed the whole, i.e., a collective reading.

⁴ There is little work to our knowledge that directly examines differences between joining and separating events; one exception is Regier and Zheng (2007), who found that joining verbs covered a smaller range of exemplars (had narrower “semantic breadth”) than separating verbs. We return to this in the general discussion.

⁵ We did not directly compare Tight Symmetrical and Tight Asymmetrical events to each other in our analysis of the four event types. Since these two events differ only on the factor of Motion, models that include Motion as an individual factor would account for any differences between Tight Symmetrical and Tight Asymmetrical events.

⁶ We also fit models with interactions between the three event factors (Motion, Fit, Direction), but there were no significant interactions, and a likelihood ratio test showed that the best fitting model was one with no interactions between the event factors (see Table S11 in Supplementary Information).

⁷ We also ran models with interactions between Age Group, Event Type and Direction, but including the interactions did not improve model fit according to a likelihood ratio test (see Table S12 in Supplementary Information).

⁸ We also ran models with interactions between Age Group, Motion, Fit, and Direction, but there were no significant interactions and a likelihood ratio test showed that the best fitting model was one with no interactions between the event factors (see Table S13 in Supplementary Information).

⁹ We also ran models that included the interaction of Age Group, Event Type, and Direction, but the interactions were not significant and did not improve model fit according to a likelihood ratio test (see Table S14 in Supplementary Information).

¹⁰ We also ran models that included interactions of Age Group and Event Type and Age Group and Direction, but these interactions were not significant and did not improve model fit according to a likelihood ratio test (see Table S16 in Supplementary Information).

¹¹ We also ran models that included interactions between Age Group and the three event factors, but none of the interactions were significant, and inclusion of the interactions did not improve model fit according to a likelihood ratio test (see Table S17 in Supplementary Information).

¹² We also ran model that included Direction and the interaction of Age Group and Event Type, but these did not improve model fit according to likelihood ratio tests (see Table S18 in Supplementary Information).

¹³ It is currently debated whether such non-linguistic changes are an inevitable effect of linguistic encoding. Some have argued that any influence of language on perceptual discrimination (whether perception, post-perception, or both) constitutes an effect of language on thought (Gilbert et al., 2006); others have argued that effects of language on language must be carefully distinguished from effects of language on non-linguistic thought (L. Gleitman & Papafragou, 2012; Landau, 2022).

¹⁴ “Sometimes fit is suggested to us as the English counterpart of kkita, but fit does not fit: in one way it is too general and in another too specific. Too general because for kkita but not fit, figure and ground must have complementary shapes before the action is carried out, and the fit requires at least a slight degree of three-dimensional engagement thus, kkita cannot be used in contexts like ‘Does this belt fit?’ or ‘This bandage is too small to fit over the wound’). Too specific because fit is typically used only when the degree of fit is the point at issue, and not for actions like putting a cassette into its cassette case or the cap on a pen. Relatively low frequency English words like interlock, mesh, couple, or engage come a bit closer, but the first two suggest the involvement of more than one projecting part from each object, and the second two evoke the notion of a connecting link between two entities, such as train cars, so it is absurd to use them for putting a book into a tight box-cover or a cap on a pen – perfectly normal uses for kkita. The meaning of kkita can, of course, be approximated in English by combining words into phrases such as tight fit, as we have done in this chapter, but such phrases are inexact and cumbersome, and, as ad hoc compositions, they are not part of the permanent stock of semantic categories of English.” (Bowerman & Choi, 2003; FN2, p. 419).