$$\rightleftharpoonup{xx}$$
$$\longleftharp{xx}$$,
$$\longrightharp{xx}$$,
The relationship between language and non-linguistic representations is a fundamental topic in the cognitive sciences1,2,3,4. In exploring this relationship, we focus on spatial cognition. The memory game procedure enables us to experimentally control the influence of different parameters on the relations between spatial language, spatial memory, and object knowledge, while also retaining a degree of ecological validity. Past methods used to elicit spatial demonstratives or the comprehension of them range from those that have high ecological validity but low experimental control (e.g., the observational work of Enfield5, or the elicitation methods developed in Max Planck Institute field guides6) to those that have high experimental control but low ecological validity (such as the within-participant designs employed tapping the congruence of demonstratives with pictures7,8). The memory game method was developed not as a substitute for these methods, but rather as a complementary method retaining the strengths of these various approaches within a single paradigm.
Key to the development of the method was the desire to retain high experimental validity while also ensuring that participants use language naturalistically in (real) three-dimensional space without being aware that their language was being tested. There are several important points to note here. First, the brain systems underlying peri-personal (near) space and extra-personal (far) space in vision and action are reasonably well charted, and involve a (graded) distinction between the space reachable around the body and the space not reachable9,10,11. In past linguistic investigations of the influence of distance on demonstratives, the perceptual basis of this distance distinction is often not adequately considered. The use of photographs in some past studies manipulating distance where the whole image on the screen is in peripersonal space is arguably not a fair test of the influence of distance on demonstratives as motivated from the basic brain system distinction. Second, asking participants to produce demonstratives and telling them that the researchers are interested in the demonstratives used opens up the possibility of bias, with participants generating their own theories regarding demonstratives and thus not producing them naturally. For that reason, the memory game uses a cover story to elicit demonstratives without participants realizing that the demonstratives chosen are of interest. Indeed, on debriefing we find that participants report being unaware of the real purpose of the study. Moreover when the purpose of the study is revealed, participants often describe how they use demonstratives in ways that do not necessary accord with their own actual behavior during the task.
There are two basic versions of the memory game, exploring language use (from here the 'language version') and object location memory (from here the 'memory version'), in which we can manipulate different parameters (see section 3.3). In the context of questioning research findings exploring top-down effects of cognition on perception12, the memory game aims to avoid the pitfalls identified by Firestone and Scholl, such as overly confirmatory research designs (different models are tested, allowing disconfimation) and demand and response biases (cover stories ensure participants are unaware of the aim of the study). (See attachment for a transcript of the instruction for both versions of the memory game.)
In the language version (section 3.1) of the memory game, testing spatial language production, we use memory as a cover story so that demonstratives can be elicited without participants realizing that their use is being measured. Participants are instructed that they are taking part in an experiment examining the influence of language on memory for object location (the experiment is advertised as a memory experiment). Participants sit at a long table with a number of color-coded locations marked at various distances in front of them. At the start of each individual trial, the experimenter or the participant (agent is a potential experimentally controlled parameter) places an object (e.g., blue heart, black cross) at one of the locations. Between trials, the distance from a participant is varied, as well as other parameters of potential interest, such as ownership (whether the participant owns an object or not), visibility, familiarity, agent (who places the object), and the position of a conspecific. After object placement, participants point at the object (but do not touch it) and name it. Participants are instructed to use three words: demonstrative, color, object name (e.g., for the English version: "this/that red circle") .
Figure 1. Overview of the table setup and positions of speaker (participant) and hearer (experimenter). Adjusted from Coventry et al.1 Please click here to view a larger version of this figure.
Figure 2. The participant reads out the instruction card, then memorizes the object location and finally instructs the experimenter to move the indication stick and align it with where the edge of the object was14. Please click here to view a larger version of this figure.
In the second version of the memory game (section 3.2), we test memory for object location. Following object placement, participants view the object/location for 10 seconds. After the 10 seconds, the object and location markers are removed, and participants verbally instruct the movement of an indication stick (see Figure 3, the indication stick is placed at various distances, either closer or farther from the actual location) to match the exact location they thought the near edge of the object was at. The experimenter moving the indication stick stands behind a curtain to avoid a clever Hans effect (i.e., so that the participant cannot read any accuracy clues from the experimenters' face/body language). The difference between the recalled location and the actual object location shows the direction and strength of the influence of the respective conditions (see Figure 4).
Across multiple series of experiments employing these procedures, we found a close relation between spatial language and spatial memory [e.g., see Table 1]. Distance and multiple parameters of object knowledge (e.g., ownership, visibility, familiarity) have been found to influence the use of demonstratives (this/that)13. For example, if an object is out of reach, one is more likely to say that compared to this1,13; if the object placed is owned by the participant, one is more likely to use this compared to when the placed object is owned by someone else. Furthermore, results from the memory version parallel results from the language production version. In situations in which participants are more likely to refer to an object using that compared to this, participants misremember the object to be further away in the memory version13.This effect is also extended to language at instruction affecting spatial memory: if the object is placed with the word that instead of this (e.g., if the participant reads out an instruction for object placement: "Place that object on the location"), participants misremember the object to be further away in the memory version14. More specifically, the influence that object knowledge has on spatial language and memory (e.g., objects placed further away are verbalized using the demonstrative that rather than this) is similar to the influence that spatial language has on spatial memory (objects placed with that are misremembered to be further away than placed with this). This shows a close relation between spatial cognition and language.
Theoretically these methods have been used to differentiate between several possible models predicting different influences of spatial language on spatial memory. For example, the Expectation model13 suggests that spatial memory is a concatenation of the expected location and the actual location. For instance, if one owns an object, it is expected that the object will be closer than if the objects belongs to someone else (since most objects that are close by are objects one owns). The expectation model is based on previous results1,13, and the evidence across experiments favors this model over other models in follow-up studies14.
