Conceptual Transfer in Second—Language Acquisition: A Case Study of Preposition Use by Japanese Learners of English
THESIS by Rachel Kamins
Prepositions are widely considered among the most difficult aspects of English for adult second- language learners to acquire. This difficulty arises from their ambiguous nature: they are highly polysemous, and they can serve two syntactic functions—either as lexical words, contributing semantic content to sentences, or as grammatical functors, helping to hold together sentence structure. They are inherently phraseological elements; their meaning becomes clear from the words with which they combine. For example, consider the sentences the book is on the table and the book is on fire; in the first example, acting as a lexical word, the preposition on specifies the location of the book, while in the second, as a grammatical word, it predicates fire as the condition of the book.
The literature on phraseology in L2 acquisition (e.g., Howarth 1998; Nesselhauf 2005; Conklin and Schmitt 2008; Paquot and Granger 2012) distinguishes several types of phrases in which prepositions can appear. An important distinction for the purposes of the present research is between free combinations, on the one hand, and restricted collocations or idioms, on the other. In free combinations, all of the combined elements carry semantic content, are interpreted literally, and can be substituted by other lexical items of the same category, subject to the semantic constraints of the language and the real-world properties being described. For instance, in English we can say on the table, under the table, and over the table for most tables that we encounter, but we can say in the table only for tables that have an interior volume (such as a drawer under the tabletop). In collocations and idioms, one or more elements or the entire phrase has a figurative meaning and cannot be substituted—in English, we can say in danger and at risk but never at danger or in risk—and the constraints on their combination are arbitrary, not semantic. More specifically, when a preposition occurs in a restricted collocation, such as attack on—that is, when it occurs in the complement to a noun that has a selection restriction for that preposition—the preposition is considered to be specified in the entry for that noun in the mental lexicon. When it occurs in a fixed idiom such as on the other hand, the whole idiom is considered to be a single lexical entry. But when it occurs in a free combination such as on the table, on and table are considered separate lexical items that are combined on the basis of real- world conditions.
Research on L2 learners’ phraseological production has found ample evidence that prepositions in restricted collocations are problematic for learners (e.g., Howarth 1998; Nesselhauf 2005; Paquot and Granger 2012). Some of these authors have indicated that free combinations pose less of a challenge because learners can rely on semantic reasoning to formulate them, while they can only memorize the many selection restrictions governing restricted collocations, which are likely to vary in unpredictable ways from similar restrictions in their native language. However, other findings indicate that restricted collocations offer their own comparative advantages; Conklin and Schmitt (2008) suggest that this is due to their frequency and salience, which allow them to be memorized relatively easily (see also Yuan 2014). Free combinations of prepositions with nouns may cause trouble precisely because they depend on semantic reasoning—and semantics are not neatly aligned across languages.
The idiomaticity of lexical prepositions has been previously noted. In an analysis of texts translated between English and French, Cosme and Gilquin (2008) find that prepositions that are commonly considered each other’s translation equivalents (with and avec) have mismatched ranges of extension in both their free and “bound” (restricted) uses. Moreover, free uses of with are less likely than bound uses to be translated as avec. The authors note that in the case of French and English, bound uses of prepositions, especially in high-frequency phrases, are often equivalent across the languages, while “many free uses are in fact not directly translatable into learners’ mother tongues” (p. 271). Shinkawa (1979), in a comparison of English prepositions with Japanese postpositions in locative expressions, notes that both of these types of spatial words, though they have lexical content, are constrained by selection restrictions in a sense. While Japanese postpositions are selected for by the verb, English prepositions are selected for by the noun. An English learner “must know the semantics of each noun before he can decide which preposition will accompany it” (p. 176). Similarly, Herskovits (1986) points out that the choice of a lexical preposition in English locative expressions is in many cases a matter of idiomatic convention, even if it is not perceived as such by native speakers.
Thus, relying on semantic reasoning to select a preposition to use in a free combination may not be a simple matter for learners of English as a second language (L2). The present study looks at preposition use in English writing by L2 learners whose native language (L1) is Japanese, seeking evidence that differences in the semantics of locative expressions in the two languages cause errors in learners’ choice of spatial prepositions. A mechanism that could explain non-native-like production of English spatial prepositions by L2 learners is conceptual transfer (Jarvis 2007; Jarvis and Pavlenko 2008), in which learners are hypothesized to keep concepts shaped by their L1 intact even while using the lexical and grammatical resources of the L2. Under the theory of linguistic relativity, differences between languages cause their respective speakers to form different conceptualizations of real-world properties such as spatial relations; thus, conceptual transfer would entail learners’ applying English vocabulary and grammar to concepts of space shaped by Japanese.
This question has received little attention in the L2 acquisition literature to date. Among recent studies that focus on L2 English learners’ use of prepositions (Yuan 2014; Mahmoodzadeh 2012; Jiménez Catalan 1996), none systematically differentiates between prepositions occurring in free combinations and those in restricted collocations or idioms. This study contributes to filling this gap by examining the use of lexical prepositions in topological spatial descriptions, in which the location of a static object is described in relation to a reference object. I seek evidence of transfer errors in such descriptions in essays written by 365 Japanese learners with an average intermediate level of English proficiency, collected in the International Corpus of Learner English (ICLE). I find only a little such evidence. The Japanese learners’ English production, while in many ways unlike that of native English speakers, appears to contain a relatively small proportion of preposition errors that can be attributed to conceptual transfer. This may indicate that conceptual transfer is not a strong explanation for L2 learners’ difficulty with prepositions, though more data are needed to fully consider the hypothesis.
The paper proceeds as follows. First, I review the literature on the causes of non-native- like L2 production, including conceptual transfer, as well as the theory of linguistic relativity in the domain of spatial relations. In the next section I provide comparative descriptions of the spatial semantics that pertain to the preposition on in English and to various terms in Japanese, present my hypotheses on conceptual transfer by Japanese learners of English in their L2 descriptions of ON relations,  and describe the data used in this study. I then lay out my findings, which are preliminary on account of a lack of relevant data in the ICLE and the limited scope of this study. I discuss the possible implications of the current findings, and finally conclude by setting out directions for further research to better determine the extent of conceptual transfer in the domain of spatial language.
L2 Acquisition and Transfer Effects
Ellis’s (2006a,b) associative learning framework for L2 acquisition outlines the general mechanisms that can stymy native-like fluency in L2 learners. Normal acquisition of an L1 is a process of contingency learning, in which the child learner intuitively and subconsciously computes the statistical likelihood of form-to-meaning mappings based on input from other speakers. In L2 acquisition by adult learners, the existence of the L1 map influences the learners’ attention to linguistic cues and interpretation outcomes, in effect inducing them to attempt to learn the L2 as if it were the L1. For any L1 or L2 learner, the lack of a one-to-one mapping between cues and outcomes—that is, ambiguity in the meaning of words—poses challenges for learning (see Degani, Tseng, and Tokowicz  on the difficulty of ambiguity for L2 learners in particular).
Prepositions are a classic case of ambiguity: “Consider the range of meanings of the English preposition in, as illustrated in the story of the man who was wounded six times during World War I, the first time in Belgium, the second in the morning, the third in the leg, the fourth in his sleep, the fifth in August, and the sixth inadvertently” (Ellis 2006b, p. 167). While native and non-native speakers alike are confronted with this ambiguity, the latter may additionally be subject to interference and overshadowing from the existing associations between similar words and meanings in their L1. Interference over time can lead to “blocking,” inhibiting the formation of associations in the L2. L1 form-to-meaning mappings can have a pervasive influence on learners’ L2 production as well as their comprehension (Trenkic, Mirkovic, and Altmann 2014). 
This influence commonly takes the form of transfer, whereby characteristics of the L1 are applied by the learner to the L2. This can take place at any level of language, from the phonological to the discursive (Jarvis and Pavlenko 2008). Jarvis and Pavlenko (2008; see also Jarvis 2007) argue that in addition to formal elements of language, underlying concepts can be transferred as well. As set out by Jarvis (2007, p. 52), the conceptual transfer hypothesis (CTH) describes “how L2 users make use of lexicalized and grammaticized concepts acquired through one language while performing in another language, as evidenced through the ways they categorize, label, frame, recall, and refer to objects, events, qualities, patterns, relationships, emotions, and so forth in the other language.” In an earlier work, Pavlenko (2002, p. 76) called this “subordinate” bilingualism, “when the conceptual system of the L1 underlies both lexicons or, in other words, when the meaning for the L2 is supplied by the L1.”
Many factors may determine when a learner relies on transfer. For instance, transfer is more likely when the learner perceives greater similarity between the L1 and L2 on a typological level or on the level of specific structures. Greater attention to and awareness of language use by the learner can lower the likelihood of transfer, whereas conditions that constrain attention and awareness can increase the likelihood (Jarvis and Pavlenko 2008). With respect to conceptual transfer in particular, Jarvis and Pavlenko (2008, p. 144) note that when a concept encompasses both prototypical and peripheral category members, transfer of peripheral concepts is more likely to be negative, as prototypical concepts are more likely to be congruent across languages. Finally, learners’ L2 proficiency does not perfectly correlate with across-the-board decreases in transfer, but as learners gain more L2 experience their performance in most areas (aside from phonology) tends to become more native-like.
Evidence of conceptual transfer, in the case of spatial semantics, would manifest as an attempt by an L2 learner to adapt the lexical and grammatical resources of the L2 in order to carry over her L1 conceptualization of spatial relations. This could result in varying degrees of ungrammaticality or infelicity in the L2.  When conceptual transfer is accompanied by lexical or syntactic transfer, the result may be words and grammatical constructions used in a non-native- like manner—for example, Reshöft and Gralla (2013) found that L1 German learners of English had a different frequency distribution of prepositions than native speakers and a non-native-like tendency to give static rather than dynamic descriptions of events. Alternatively, the transfer may not involve formal differences from typical native production, but rather non-conformance to the native mapping between words and meaning—an example that we might expect from L1 Japanese learners (as explained in more detail below) is “The cat is sitting over the mat,” where the conceptual distinction in English between on and over is not observed. Last, where the L1 and L2 are conceptually congruent, conceptual transfer may result in no infelicity at all. Given that individual speakers of the same language can vary in their mode of expression to a certain degree, evidence of conceptual transfer is expected to manifest in the form of statistical tendencies that differentiate non-native from native speakers, not as a categorical divide between the two (Jarvis 2007).
Linguistic Relativity and Spatial Language
In recent decades, a spate of research has provided evidence for the theory of linguistic relativity—the theory that the language a person speaks affects the concepts she forms. Much work on differences across languages with respect to spatial concepts has been done by Melissa Bowerman, Stephen Levinson, and their colleagues and co-authors. This work has focused on spatial distinctions that are conventionally marked in some languages but not in others. While any given language can be manipulated to describe any aspect of a spatial relation, linguistic relativity is tied to the distinctions made (or not made) in the course of a typical, standard spatial description. For instance, the English language conventionally distinguishes between objects that are in a containment relation (where one object is said to be in the other) and objects that are in contact and support relations (where one object is on the other). Dutch makes finer distinctions within the ON category: it differentiates among contact relations that involve horizontal support; vertical support where one object is attached to the other; vertical support where the objects are not attached; and support where one object encircles the other. Like English, Dutch also has a separate category for containment relations. Other languages, however, do not: in Korean, there is typically no distinction between containment and support, but instead an obligatory distinction between relations in which objects fit tightly together and those in which they fit loosely.
In experimental work, Bowerman and Choi (2001, 2003) studied how these cross- linguistic differences play out in children’s and adults’ production and comprehension of spatial language. They found that children learn very early to attend to the aspects of spatial relations that are conventionally distinguished in their language —and that adults lose some of their ability to categorize certain aspects of spatial relations that their language does not distinguish.
Similarly, McDonough, Choi, and Mandler (2003) found that English-speaking adults typically did not note a difference between tight-fit and loose-fit relations when asked to find a difference among relevant spatial scenes, whereas Korean-speaking adults made the distinction. The Korean group’s performance reflected the categories of its native language, whereas the English group’s performance reflected its lack of relevant linguistic categories.
In addition to these experimental findings, semantic typological work on spatial language (e.g., Bowerman and Pederson 1992; Levinson et al. 2003; Levinson and Wilkins 2006; Zhang, Segalowitz, and Gatbonton 2011) has been argued to show that there are no universal basic concepts (“primes”) in spatial relations. By eliciting descriptions of spatial configurations from speakers of a wide range of languages and comparing the semantic categories that emerge within each language, these studies have found evidence of pervasive cross-linguistic differences in theconceptualization of space.  For instance, Levinson et al. (2003) compared descriptions of pictures showing IN and ON relations across a dozen languages and found no agreement between any two languages over the grouping of any two pictures. Levinson and Wilkins (2006, p. 551) conclude, given the lack of concrete, universal spatial primes, that “the child language learner is a constructivist—he or she is not just mapping local forms onto pre-existing innate concepts but building those concepts as he or she learns the language.” Thus, the influence of the child’s language is built into her concepts.
Jarvis and Pavlenko (2008) hypothesize that L2 learning of spatial language involves transformation of spatial conceptual categories. Learners must reorganize their conceptual map of spatial relations, consolidating some concepts with disparate semantics in their L1 and segregating others that have shared L1 semantics, and then appropriately linking the new semantic categories with L2 vocabulary and syntactic frames. Moreover, they must redirect their attention to spatial relations in order to newly distinguish aspects that are not habitually linguistically encoded in the L1, while disregarding some distinctions that are not habitually marked in the L2. To the extent that these processes of reorganization and redirection remain incomplete at a given stage of their L2 development, learners may be susceptible to conceptual transfer.
The Case Study: ON Relations in English and Japanese
Japanese varies notably from English in its categorization of ON relations; it has no single category of spatial relations corresponding to the English on category, nor a single preposition (or functionally similar word) with the same range of extension as on, nor an identical syntactic form for expressions describing ON relations. This section will delineate both the semantic categories of Japanese that correspond to English ON relations and the morphosyntactic encoding of these semantics. 
Before proceeding, I will set the scope of this research project and define some of the terms to be used below. As noted in the introduction, this study focuses on descriptions of topological relations. Following a tradition stretching back to Talmy (1978), to which Herskovits (1986) contributed, Levinson and Wilkins (2006) use this term to refer to relations in which two objects, a figure and a ground, coincide in space. The figure is a movable or “conceptually movable” object (Talmy 1978, cited in Herskovits 1986, p. 35), though in stasis at the moment of description, and the ground is a stationary reference point. Their coincidence may entail, more specifically, propinquity, contact, or containment. The comparative typology of topological spatial relations in Levinson and Wilkins (2006) is based on descriptions of the 71-item Topological Relations Picture Series (TRPS; Bowerman and Pederson 1992), elicited from native speakers of 11 different languages, including Japanese (described by Kita ).  Each picture in the TRPS shows a figure object in some topological configuration with a ground object; informants were asked, “Where is the [figure]?” and desirable answers provided the location of the figure with respect to the ground.
The grammatical resources used to provide the location information vary widely across languages—this is a major finding of Levinson et al. (2003) and Levinson and Wilkins (2006). Whereas in English the location is primarily encoded in prepositions (e.g., “The cup is on the table”) other languages make use of postpositions, case-marking, and/or richly informative locative verbs. Another part of speech, used in Japanese, is the spatial nominal, a type of noun that refers to a specific region of a ground object and is used in conjunction with the Japanese locative postposition. These are analogous to the English nouns top, inside, bottom, and the like. Finally, the basic locative construction (BLC) of each language is the syntactic frame in which its speakers provide locative information for the most prototypical topological relations. The BLC may be extended to less canonical variations on topological relations, or those variations may evoke different constructions. Levinson and Wilkins (2006) provide a hierarchy of spatial scenes in terms of their cross-linguistic likelihood of being described by a BLC, reproduced here as figure 1.
English ON Relations
The ON category in English describes relations of contact, often but not necessarily with support. Herskovits (1986, p. 140) defines the “ideal meaning” of on as follows: “for a geometrical construct X to be contiguous with a line or surface Y; if Y is the surface of an object OY, and X is the space occupied by another object OX, for OY to support OX.” In actual usage, this meaning is subject to an array of extensions, transformations, and shifts. Often the geometric nature of the figure and the ground is a conceptualization—an imposed mental construct. It is also often the case that when the prototypical geometry of contact relations does not apply, a support function between the figure and ground warrants the use of on, as when we say that the top book on a pile is “on” the table, even though the table is only contiguous with the bottom book in the pile (see also Garrod, Ferrier and Campbell  on functional definitions of spatial relations).
ON relations encompass almost the full range of situations in the BLC hierarchy, except for item (2), damage or negative space, which is more typically encoded with in. A small, movable figure may be supported by a ground with which it is in contact (e.g., a book on a shelf; Picture 8 in the TRPS [henceforth, TRPS 8]); the figure may be worn as an adornment or clothing (e.g., a ring on a finger; TRPS 10); it may be part of a whole, where the whole is the ground (e.g., a strap on a bag [TRPS 66] or “the wrinkles on his forehead” [Herskovits 1986]); it may be stuck to the ground (e.g., a stamp on an envelope; TRPS 3); or it may be impaled by the ground (e.g., papers on a spike; TRPS 22). The figure or the ground may be animate, either human or animal (e.g., a person on a roof, a bug on a ceiling, and a necklace on a neck [TRPS 34, 7, and 51]).
Some other examples of extended and sense-shifted uses of on, drawn from Herskovits (1986, pp. 140–48) include the following: the ground may be a large vehicle that is transporting the figure (as in “the children on the bus”); the figure may not be a material object (as in “the image on the movie screen”); the ground may be a part of the figure that supports the figure (as in “a table on three legs”); the ground may be a geographical location (as in “the biology lab is on the campus”) or a physical or imaginary line imposed on geography (as in “Is Lima on the equator?”); or the figure may be on the edge of the ground (as in “the store on Polk Street”). This variety of situations described by the single English preposition on is remarkable when compared to the distinctions carved through this semantic space by other languages (see, e.g., the cross- linguistic comparison in appendix 1 of Levinson and Wilkins ).
The locative constructions that will be considered in this study contain an explicitly named figure and a prepositional phrase in which the ground is the object of the preposition. The English BLC takes the form Figure_NP Copula [P Ground_NP],  as in “The figure is on the ground.” An existential variation on the BLC occurs frequently: Expletive_Subject Copula Figure_NP [P Ground_NP], as in “There is a figure on the ground.” The figure NP and the PP containing the ground may also be combined in a variety of ways beyond the BLC, with the preposition continuing to convey the primary information about the spatial relationship between the figure and ground. For instance, while copular be is the most typical locative verb in English, more-specific positional verbs may be used for emphasis, as in “The figure is standing/sitting/lying on the ground.” Other verbs that connote the existence or stasis of the figure with respect to the ground (e.g., the figure exists, lives, stays on the ground) or observation of the figure in relation with the ground (e.g., someone sees, finds, locates the figure on the ground) are also appropriate to the present analysis.  I will further include relative clauses in which a trace of the moved figure NP directly c-commands the PP (e.g., “[The figure]i which ti is on the ground”).
However, I will exclude from consideration expressions in which the figure NP and the PP containing the ground are both complements to a finite verb of motion, as in “I put the figure on the ground.” As shown by Slobin’s body of work (e.g., 1996a,b, 2003) and Levinson and Wilkins (2006), motion expressions are complex constructions that may convey multiple pieces of locative information, including the source, path, and goal of motion, in addition to information about the manner or aspect of motion. Such considerations are beyond the scope of this study, which focuses on simple topological location as depicted in the TRPS and as described by Herskovits (1986). However, I will include expressions in which a past participle intervenes between the figure and the ground as a modifier of the figure, in the form of a full or reduced relative clause—for instance, “The figure (that/which is) written, projected, displayed on the ground”; these can be considered to refer to a static end-state resulting from motion, rather than to the dynamic act itself.
Japanese ON Relations
As described by Kita (2006), the Japanese BLC is used only to describe a limited range of situations, largely encapsulated by item 1 in the BLC hierarchy: those in which the figure is relatively small and movable and is vertically supported by a relatively large and stable ground; the figure is not attached to the ground and could move freely in at least two dimensions; or the figure is inherently fixed with respect to the ground.  Its morphosyntactic form is Figure_NP-wa Ground_NP-ni Locative_VP, where the postposition -ni appended to the ground NP assigns the dative case and conveys a generic sense of static location; in expressions of location, it is commonly translated as at as well as in and on.  This construction expresses the notion “[the figure] is at/in/on [the ground].” Some examples of expressions that would take this construction are Yamada-san-wa Tokyo-ni i-ru, ‘Mr. Yamada is in Tokyo,’ and Kyomizudera-wa Kyoto-ni a- ru, ‘Kiyomizu Temple is in Kyoto’ (Kita 2006, exx. [2a,b]). The verb in the BLC can take one of four forms depending on the animacy of the figure and honorificity, but in any form it essentially conveys the sense ‘is.’ A variation on the BLC with similar meaning is the existential construction: Place_NP-ni Figure-NP-ga Locative_VP, “at/in/on the ground there is the figure.”
The dative case particle –ni is in complementary distribution with the form -de, which, in the realm of spatial relations, has the same meaning and function as –ni but is used to describe the location of a dynamic activity rather than a static object (Martin 1975; Shinkawa 1979). Some verbs other than the copula can be used with –ni, especially verbs of stasis (taizai suru, ‘stay,’ tomaru, ‘stop’) and verbs of existence (sumu, ‘live’). Certain adjectival and nominal forms evoke the use of –ni, though most adjectival forms evoke -de. Some verbs that denote activities but not necessarily motion, such as “work,” may also evoke –de. The choice of –ni or – de depending on the desired emphasis (Shinkawa 1979). The division between relations that are static and those that are dynamic is not a bright line; “The choice of ni or de is often the only indication that stasis is present” (Martin 1975, p. 218). Henceforth I will refer to this spatial particle as –ni/-de to allow for this unpredictable variation.
A third form of the Japanese BLC includes a complex ground NP used to specify a particular region of the ground where the figure is located. Relevant to ON relations, the spatial nominal ue on its own means ‘top’; when combined with the postposition -ni, it refers to vertical relations between a figure and a ground.  This form of the BLC, Figure_NP-wa Ground_NP-no ue-ni Locative_VP, is roughly equivalent to “the figure is at the upper region of the ground” (the ground NP is in the genitive case, indicating that the “upper region” referred to by ue belongs to it). A crucial difference between Japanese and English for the purposes of this study is that ue, like all Japanese spatial nominals (Kita 2006, p. 446), does not imply contact, so it conflates ON and OVER/ABOVE relations.  Thus, a lamp sitting on the surface of a table (most closely analogous to TRPS 1—cup on table) and a lamp hanging from the ceiling above a table (TRPS 13) are both described with ue in Japanese. 
A range of other types of ON relations are expressed with resultative or imperfective locative verbs, which indicate that the location of the figure is an end-state resulting from some action. These constructions are strongly preferred whenever it is typically the case, likely, or even merely conceivable that the figure has been put in its location through some preceding event, whether or not the speaker has any direct knowledge of that event. To save space, I will not reproduce the morphosyntactic frame of each such expression, but only provide glosses (for details, see Kita 2006, pp. 441–45). Such relations include those of a stamp on an envelope (TRPS 3), which is described by a construction that glosses as “The stamp is in the state of having been stuck to the envelope”; a handle on a cabinet door (TRPS 61), “The handle is in the state of having attached to the door”; an apple on a skewer (TRPS 70), “The apple, a skewer has pierced”; a ring on a person’s finger (TRPS 10), “As for that ring, Mr. Yamada is wearing [it]”; or a ring on an inanimate object (analogous to TRPS 4—a ribbon on a candle), “The ring is in the state of [being/ having been put] into a tight-fit configuration with [the stick/the mummy’s middle finger].” 
As described in the preceding sections, the category of spatial relations described by the preposition on in English maps to a number of separate semantic categories in Japanese, and these Japanese categories in turn map back to other English words in addition to on. The hypotheses of this study reflect the asymmetry of this mapping, predicting that evidence of conceptual transfer will occur in the form of conflation of semantically distinct English words that correspond to single categories in Japanese.
First we consider the category of vertical ON relations between a figure and a specific region of a ground, described by the spatial nominal ue in Japanese. Ue maps to both on and over/above in English. Hence, I will look for evidence of mutual substitution of these prepositions, as stated in hypothesis 1.
Hypothesis 1. Uses of on and over/above by Japanese learners of English will be likely to show conflation of the semantics assigned to these prepositions in English, especially in expressions that would use the spatial nominal ue if construed in Japanese.
A second category of ON relations is that of canonical relations of spatial coincidence between a small, movable or inherently fixed figure and a relatively large and stable ground, described by –ni/-de in Japanese. These Japanese particles in turn map to on but also to at, in, and other prepositions in English. Hence, I will seek evidence of mutual substitution among these prepositions as well, searching in learners’ uses of the two prepositions on and at.
Hypothesis 2. Uses of on and at by Japanese learners of English will be likely to show conflation of the semantics assigned to these prepositions as well as other prepositions in English, especially in expressions that would use the spatial particle –ni/-de if construed in Japanese.
Jarvis and Pavlenko (2008) set out three criteria for identifying “cross-linguistic influence,” or transfer, in a population of language users. The first is intragroup homogeneity: the population should have a relatively uniform level of proficiency in the source and recipient
languages (in this case, the L1 and L2). The second is intergroup heterogeneity: the population’s language use should be significantly different from that of another group (in this case, native English speakers). Third, the population’s language use should demonstrate cross-linguistic performance congruity; this criterion would be met in the present case by a preponderance of evidence for the hypotheses set out above, which would also satisfy the criterion of intergroup heterogeneity. Intragroup homogeneity is partly satisfied by the fact that all learners in the study population are native speakers of Japanese, with relatively little experience, on average, with other foreign languages. These characteristics and other considerations for determining intragroup homogeneity will be discussed further in the following subsection, where the ICLE data are described in detail.
The data used in this study are collected in the International Corpus of Learner English (ICLE), version 2, compiled at the Centre for English Corpus Linguistics at the Université Catholique de Louvain.  For every contribution to the corpus a questionnaire was completed detailing several characteristics of the writer and the conditions under which the essay was composed; these metadata are encoded in the database. Further, each word (or multiword unit) in the corpus was tagged with its part of speech (POS) by the automated tagger CLAWS7, and the database comes with a built-in Unitex concordancer.
The present study uses the L1 Japanese subcorpus of the ICLE, which contains texts by 366 contributors. In order to eliminate contributors whose use of English might be affected by extensive experience with any language other than Japanese, I set the following parameters when searching the subcorpus: country—Japan; native language—Japanese; first language at home— Japanese; second language at home—none. This eliminated only one text from the subcorpus, yielding a sample of 365 texts comprising 197,251 words.  All of these texts are argumentative essays. Other characteristics of the sample are shown in table 1.
The intragroup homogeneity of the contributors can be considered from multiple angles. They are highly homogeneous in terms of linguistic experience, gender, and age, being largely 17–25-year-old females who have studied only Japanese and English. Thus, their native- language ability is likely to be consistent.  Their length of experience with English instruction is somewhat more variable; however, there are reasons to believe that this may not entail variable proficiency.  Johnson and Newport (1989) found that immersive exposure to a second language before puberty was the only significant correlate of L2 performance; the amount of formal instruction in English in the home country was not predictive of proficiency. In that regard, the ICLE data are somewhat murky, since the number of months spent in an English-speaking country is unknown for over half of the sample.
Finally, table 1 shows that the median length of the essays in this sample is in the range of 500–1,000 words; those outside this range are shorter. Biber (1990) found that text samples 1,000 words in length can be adequately representative of linguistic characteristics, particularly for high-frequency phenomena such as prepositions. It will be appropriate to interpret the results of the present analysis with some caution given the shorter length of the text samples, and also in light of the fact that the essays all represent one text genre, argumentative writing.
This subsection details the methods I used to search for and evaluate uses of the English prepositions on, over, above, and at in the Japanese learner sample. I use on as an example in the following description. I first searched for occurrences of the word on tagged as a preposition by entering the query <on.II>. I performed a number of additional checks to ensure that this search was not missing any relevant occurrences of on due to inaccurate POS-tagging.  First, I searched the sample for occurrences of on tagged as a prepositional adverb or particle (<on.RP>). The majority of these occurrences were indeed in adverbial expressions or phrasal verbs, though a few were in idiomatic expressions (e.g., and so on). I counted only one of the results of this search as a topological expression: the earth … as a place for people to live on, which was also found by the <on.II> search.
Second, I queried the sample for occurrences of on within tagged multiword expressions (<CDIC><<on>>). Many of the results of this search were idiomatic phrases such as and so on and from now on. A few were either literal or metaphorical spatial expressions (clinging its jaws on to the padded arm, the people who went on board the Titanic, and the emperor is on top of all of Japanese), but none were true topological spatial descriptions. Finally, I selected a sample of 20 essays from throughout the corpus and read them in their entirety to check whether topological uses of on had escaped the other queries. I found no relevant uses that had not been included in the search results, so I am fairly confident that the results I report below are exhaustive.
Within the <on.II> results, I reviewed the lines of the concordance and selected those occurrences that represented topological relations between an explicitly named figure and a ground, in conformance with the examples provided by the TRPS, Levinson and Wilkins (2006), and Herskovits (1986). This required eliminating many other types of uses of the preposition; in an attempt to improve the reliability of this manual, subjective sorting procedure, I performed it multiple times. Some common types of expression that I excluded are described below (examples are reproduced with minimal context and include original errors).
- Spatial relations involving motion or action: As noted above, topological relations involve a static relation between a figure and a ground. Therefore, dynamic spatial expressions involving either activity or motion were eliminated from the results, including, for example, construct buildings on their own lands, put a collar on him, and write my address on the envelope.
- Restricted collocations and idioms: Many of the occurrences of on followed verbs which restrictively select for on, as in based on and depend on, or appeared in fixed idioms such as on the other hand and on the contrary.
- Purely metaphorical expressions: Expressions that were locative but purely imaginary were excluded—for example, stain on its reputation and the divorce left many deep scars on both of them.
- On + communication medium: Another large group of occurrences were descriptions of communication via electronic media: on TV, on the phone, on email, on the internet, and so forth. Communication over distance does involve space in a sense, but these expressions did not refer to the topological location of a physical figure, and so they were not included.
- Time expressions: For example, on Sunday, on May 8.
I consulted with a native speaker of Japanese to determine the spatial term that would be used in Japanese to primarily convey the spatial relation described in each English example. Finally, I note that the results reported below are largely qualitative, though I include raw and normalized frequencies for all uses of the prepositions and for errors, as well as the percentages of texts in which occurrences and errors are dispersed. These numbers aid in assessing the prevalence of conceptual transfer and will also provide a basis of comparison for future studies.
The <on.II> search yielded 727 occurrences of prepositional on in 278 texts (76.2% of all the texts in the sample; the frequency of occurrence of on per 100,000 words is 366.7). After the sorting procedure described above, there remained 47 topological spatial expressions using the preposition on (23.7 per 100,000), which occur in 36 texts (i.e., they represent the production of36 different learners, and 10% of texts). These are given in table 2; I have provided some corrections and clarifications in square brackets. The right-hand column indicates the spatial nominal, particle, or verb that would be used to convey the primary information about each spatial relation if the sentence were cast in Japanese.
As can be seen by reading through the “usage” column, the English of the Japanese writers using topological on is hardly error-free. Yet the use of the preposition on in topological expressions is rarely in error. The only 2 errors among the 47 examples involve conflation of on and in: one thing written on the newspaper and there were many dogs left free on the alleys in Bangkok (exx. 3 and 13). In the context of essay HI-0003.1 (ex. 3), it is clear that the writer is referring to her ability to understand the published text of newspapers, which native English speakers would typically refer to as text “printed in” newspapers. “Written on” would be appropriate for handwriting added to an already-printed newspaper. The spatial relationship of dogs in alleys, referenced in essay KO-0026.2 (ex. 13), belongs to a use type that Herskovits (1986, p. 154) assigns to in: “physical object in roadway.”
There is no apparent evidence for hypothesis 1 in these examples—that is, no potential conflation of on and over. One example, the sentence the oven was on the refrigerator (ex. 25), bears some consideration. In essay SWU-0004.2, the writer is describing the kitchen in the American home of her host mother in an exchange program. Her purpose in describing it is to relate a story told to her in a letter from her host mother. In Japanese kitchens (according to my informant), it is quite common to have a short refrigerator with a small oven sitting directly on top of it. However, in American kitchens both appliances are much larger, and this arrangement seems implausible—most adults couldn’t reach an oven positioned above a refrigerator, whether it was directly on top of it or over it, separated by cabinetry. It seems likely that the writer was misremembering both the letter and the kitchen and projecting her image of a typical Japanese kitchen onto the American one. In that case, this would not be an instance of on/over conflation but a semantically correct English description of an inaccurately conceptualized spatial relation.
This description of the oven’s location is one of the three examples for which the spatial relation would be conveyed by the spatial nominal ue in Japanese, as shown in table 2. The other two examples, he is … lying on his bed and the leading actor and actress on the stage (exx. 15, 26), contain no apparent errors that could be tied to conceptual transfer under hypothesis 1. We must also briefly consider the 9 expressions that correspond to the Japanese spatial term chikyuujou. The final character in this term is ue, but when combined with the classical Chinese characters typically used to refer to the earth, it is pronounced ‘jou.’ As a whole this term refers to topographical or geographical locations, as indicated by its correspondence to English expressions in the ICLE referring to locations on earth, on a mountain, on the moon, and so forth. The term appears not to be susceptible to translation with multiple English prepositions, being typically translated only as ‘on the earth.’ 
Turning to hypothesis 2, we find that 21 of the examples in table 2 correspond to Japanese expressions in which the particle –ni/–de would convey the primary spatial information. Two of these examples, described above, contain substitution of in for on, which could be expected if learners had transferred the conflated at/on/in semantics of –ni/-de to English. Of the14 remaining usages of on not corresponding to ue, chikyuujou, or -ni/-de, 9 correspond to expressions in Japanese that would primarily convey the spatial relationship through a spatial predicate—in these cases, either tsukeru, ‘to attach,’ or utsutteru, ‘to be reflected, projected, or displayed.’  Three of the sentences would use naka in Japanese, a spatial nominal that applies to relations of containment and is typically translated as ‘in.’  Finally, 2 of the examples would use the genitive case particle -no, typically translated as ‘of,’ to convey the spatial relationship.
I also searched the corpus for uses of over and above in order to seek evidence of conceptual transfer from ue resulting in conflation of these two prepositions with on. I found very little relevant data—4 topological uses of over (shown in table 3) and no such uses of above.  One of the over examples provides potential evidence of transfer: winter geta usually has a cover over the geta (ex. 3). A geta is a traditional Japanese wooden shoe that is like a platform sandal with thong-style straps on top. These shoes can be winterized with additional material attached to the sole to cover the toe area.  The fact that the cover is attached to the base suggests that the English preposition on would more naturally describe the relation of the cover to the geta as a whole (e.g., “there is usually a cover on winter geta”), under the use type “physical object attached to another” (Herskovits 1986, p. 144); over would be more appropriate if the cover were free of the shoe, like galoshes or spats. My informant identified this expression as corresponding to a use of ue in Japanese, making it likely that conceptual transfer is the source of this error.
Finally, I investigated whether the Japanese learners transferred their spatial semantics for ON relations to English by looking at their use of the preposition at for evidence of conflation of at with on or other prepositions. The only search that returned relevant findings was for the word at tagged as a preposition (<at.II>). This query returned 543 occurrences of at in 227 texts (62%; normalized frequency = 273.9). The 50 examples that pertain to topological spatial relations, from 33 texts (9%; normalized frequency = 25.3), are listed in table 4.
There is a much higher rate of errors among these results (13 out of 50) than among the on results, providing more potential evidence of conceptual transfer from the -ni/-de category in Japanese (hypothesis 2). Most (10) of these errors are substitutions of at for in; 6 of these 10 errors are in expressions pertaining to a school (exx. 18, 22, 23, 24, 40, 41). “At school” is permitted under the use type “spatial entity at location” (Herskovits 1986, p. 130), but only when the location is conceived of as one of a set of distinct locations where a figure could be found at a given time. For instance, if someone were looking for a person at home but found him not to be at that location, his present location could be described as at school. Otherwise, the morerelevant use type is “participant in institution” (p. 154), which can apply even when the person is not physically in the school building. All of the essays in which this error occurs are referring to a period of life by way of referring to participation in a certain school, so in seems more correct.  Of the remaining 4 substitutions of at for in, the two references to animals “at” places (exx. 13, 14) and the phrase some kimonos at dress section (ex. 19) more appropriately pertain to the use type “spatial entity in part of space or environment” (p. 152), and it is at Osaka (ex. 32) is ruled out by Herskovits’s list of geographical locations that can follow at (large cities are excluded; p. 131).
The other 3 errors among the results are all substitutions of at for on. Essay SE-0001.1 contains two descriptions (exx. 20, 21) in which baseball players are located with respect to a baseball field, a relation that Herskovits (1986, p. 147) classifies as “spatial entity located on geographical location.” Sports fields are one type of ground for which, as a matter of convention, English speakers only describe figures as being located on the ground. The third error is in the phrase someone at the opposite side of the world (ex. 47). This example does not have an obvious match among the use types that Herskovits describes for on; the closest approximation may be “physical or geometrical object contiguous with a line,” if the speaker and the subject of the prepositional phrase are envisioned as respectively located on meridian lines that are 180 degrees apart. This conceptualization could also fit a use type for at, however: “spatial entity at landmark in highlighted medium” (pp. 136–37). One example of such a landmark given by Herskovits is the equator, the (implied) highlighted medium for which is the surface of the earth. For help in resolving this ambiguity, I searched a corpus of texts by native speakers of English, the Corpus of Contemporary American English (COCA).  This is a balanced monitor corpus containing over 450 million words of written and spoken text. I searched COCA for the strings “on the opposite side of” and “at the opposite side of” and found 423 occurrences for the former combination and only 21 for the latter, a ratio of 20:1. This provides support for my own intuition that the former is more typical, and thus the usage in question from the ICLE sample can be counted as an error.
As shown in table 4, according to my native-speaker informant, all 13 of these errors, and 45 of the 50 examples altogether, correspond to expressions in Japanese in which the particle –ni/-de would convey the primary information about the spatial relationship. Hence, according to hypothesis 2, the errors are plausible cases of conceptual transfer. For the remaining 5 examples, the equivalent expressions in Japanese would primarily convey the spatial relationship through the genitive particle –no.
Summary of Results
Of the 47 topological uses of on found in the Japanese subcorpus of the ICLE, 0 are substitutions for over/above. Of the 4 examples of over, 1 is a substitution for on. Only 4 of these 51 total examples correspond to Japanese expressions that would use the spatial nominal ue; while 1 of these 4 (25% of the relevant occurrences and essays) contains conflation of on with over, this is a very small sample, and it includes no relevant data about above. There is thus only weak, suggestive evidence for hypothesis 1—judging from the few relevant expressions, intermediate- level Japanese learners of English do not appear to typically conflate the English prepositions that describe vertical relations with and without contact as these semantic categories are conflated by ue, though further investigation with a more relevant set of data is necessary.
Turning to hypothesis 2, of the 47 examples of on, 2 are substitutions for in, while of the 50 examples of at, 13 are substitutions for in or on. These 15 errors are all potentially attributable to conceptual transfer from L1 Japanese as described in hypothesis 2, as they are cases of conflation of prepositions that correspond to uses of the particle –ni/-de and thus to a single semantic category in Japanese. If we take together just the 21 expressions among the on results and the 45 among the at results that correspond to -ni/-de, we find that 23% (15/66) show evidence of the hypothesized form of negative transfer. These 15 errors occurred in 11 unique essays, or 24% of the 46 total essays that contained the 66 relevant expressions. As with the evidence on hypothesis 1, these proportions are notable but relatively small; however, confirmation or refutation of this hypothesis, too, awaits a broader set of relevant data. A key next step is extending this study to examine uses of in, another typical translation of -ni/-de and a preposition whose semantics have already been shown to be conflated with those of on and at by the Japanese learners in the ICLE.
Factors Affecting the Likelihood of Transfer
While these preliminary findings contain noteworthy proportions of errors that may relate to conceptual transfer, the majority of learners in this sample—about three-quarters in both sets of data relevant to the two hypotheses—did not commit such errors. This subsection and the following one will therefore consider possible explanations for these larger proportions of native- like English descriptions of contact and coincidence relations.
These correct uses of on and at may indicate that these intermediate-level learners, despite their other errors, had largely already learned the semantics of the relevant spatial relations in English. Jarvis and Pavlenko (2008, p. 44) note that “if … monolingual and bilingual speakers of the same recipient language exhibit similar behavior in the recipient language, then this may indicate successful recipient-language acquisition.” Some factors mentioned above may have increased their likelihood of success. For instance, transfer is more likely when a learner perceives greater similarity between the L1 and L2 on a typological level or in specific structures. To the extent that the ICLE contributors had such meta-linguistic awareness, it seems unlikely that they would have considered Japanese and English to be so similar as to warrant transfer, and thus they may have been (consciously or subconsciously) on guard against it.
Transfer errors may also be more likely to occur when learners are unable to pay explicit attention to their L2 production and are instead forced to rely on their automatic competence— that is, when they are not given the time or resources to reflect on their word choices and to correct their initial errors. Among the ICLE essays, those written under timed conditions, without the use of reference tools, or for examinations are the most likely to represent learners’ automatic competence. Let us therefore consider the writing conditions of the Japanese essays in the ICLE that do provide potential evidence of conceptual transfer. Of these 12 essays that contain errors in preposition choice, 11 are known to have been written under timed conditions, 11 were written without the use of reference materials, and 5 were written for an exam. Of all the 73 essays containing topological uses of the prepositions on and at (with or without errors), 46 were timed, 59 were written without references, and 30 were exams. Thus, 24% (11/46) of the timed essays,19% (11/59) of the essays written without references, and 17% (5/30) of the exam essays contained potential conceptual transfer errors. These proportions are similar to those reported in the main findings, suggesting that even under conditions more likely to induce transfer, the large majority of these learners demonstrated competence in the semantics of spatial prepositions in English. This suggestion could be confirmed by testing the competence of similar learners in linguistic tasks that are online to an even greater degree, such as speech production. 
The Thinking-‐for-‐Speaking Hypothesis
An interesting potential alternative to the CTH that could explain the correct performance of the majority of these learners is the thinking-for-speaking (TFS) hypothesis (see, e.g., Slobin 1996a,b, 2003). This hypothesis generates more limited predictions as compared to linguistic relativity or the CTH (Jarvis and Pavlenko 2008, p. 115), being explicitly constrained to the effects of language on thought that takes place for the purpose of production: “In the evanescent time frame of constructing utterances in discourse one fits one’s thoughts into available linguistic frames. ‘Thinking for speaking’ involves picking those characteristics of objects and events that (a) fit some conceptualization of the event, and (b) are readily encodable in the language” (Slobin 1996a, p. 76). 
As described by Gumperz and Levinson (1996, p. 26), the process of encoding thought for speech under the TFS hypothesis involves “regiment[ing] the conceptual structure so that it matches semantic specifications in the lexicon: if there is no match, the thoughts cannot be coded.” This could mean that conceptual transfer is blocked by a certain degree of semantic competence in the L2. For the majority of the Japanese contributors to the ICLE, then, it may have been the case that rather than thinking in Japanese for speaking in English, their thoughts— specifically, their conceptualizations of spatial relations—were shaped by the language they were using for the task at hand. Recent research studying the effects of linguistic relativity in balanced bilingual speakers has found that the immediate linguistic context (e.g., the language in which an experiment is conducted) determines the conceptual framework used (see esp. Athanasopoulouset al. 2015). The present findings suggest that L2 learners with sufficient proficiency may operate with similar conceptual flexibility.
On the evidence available in the ICLE and examined for this study, conceptual transfer appears likely to have some effect, though perhaps not a large one, on descriptions of spatial relations by Japanese learners of English. However, this research should be extended in several ways to further test the hypotheses. First, as noted, a key extension of the study in its present design would be to examine uses of the preposition in in the Japanese sample to look for additional evidence of conceptual transfer from -ni/-de. A second extension is suggested by Jarvis and Pavlenko’s (2008) assertion that when a concept encompasses both prototypical and peripheral category members, transfer of peripheral concepts is more likely to be negative. According to the typology of Japanese spatial semantics offered by Kita (2006), the most prototypical ON relations are those described within the BLC, with the spatial relation conveyed by -ni/-de or ue. More peripheral types of ON relations are described by spatial predicates, such as tsukeru (‘to attach’) and utsutteru (‘to be reflected, projected, or displayed’). We might also consider that the expressions found in the ICLE that correspond to uses of the spatial nominal naka (‘in’) or the case particle -no (‘of’) in Japanese are peripheral members of the ON category. Picking up errors originating from these peripheral spatial terms would require searching the ICLE for the more typical translations of these terms, to see if those translations are used by Japanese learners when on would be the more native-like choice in English. If a higher proportions of such errors were found, Jarvis and Pavlenko’s assertion would be confirmed.
Third, further research is needed that tests these learners’ competence in other contexts of production, particularly spontaneous, conversational speech and speech elicited in a controlled laboratory setting. One intriguing possibility for extending this work is eliciting descriptions of the TRPS scenes from Japanese learners of English, in their L2. This method would allow researchers to have more confidence in interpreting results by providing a clearer view of the intended referents of learners’ speech, and would also provide more relevant data by limiting the context of preposition use to topological spatial descriptions. Fourth, future work should extend the investigation of this question to other populations of English learners. For example, an L1 that is both described in the semantic typologies in Levinson and Wilkins (2006) and represented in the ICLE is Dutch; the fact that Dutch draws on four prepositions to describe ON relations would make for another interesting case study in acquisition of English prepositions. Finally, to better understand the conceptual flexibility attained by L2 learners over the course of their development, it would be interesting to replicate experiments such as those of McDonough, Choi, and Mandler (2003) and Athanasopoulos et al. (2015) with learners at various levels. Doing so might reveal how far the potential for conceptual transfer extends in the domain of spatial language.
 Following Levinson et al. (2003) and Levinson and Wilkins (2006), I use words in all capitals to represent the categories of spatial relations associated with certain English words, which may be associated with nonequivalent words in other languages.
 A model of the bilingual mental lexicon that puts more emphasis on learning than Ellis’s is the “shared (distributed) asymmetrical model” of Dong, Gui, and MacWhinney (2005). Development in this model is described as follows: “For L2 learners, L2 Name-Concept links are weaker than L1 Name-Concept links. Initially, they may be so weak that learners have to rely on lexical-level links from L2 to L1 to achieve activation of concepts. The developmental shift is, therefore, implied in the gradual strengthening of L2 Name-Concept links” (p. 234). This model also illustrates that bilinguals have a divided conceptual store, including both conceptual elements shared between the L1 and L2 and separate elements pertaining to each language. This division remains in place as appropriate links from L1 and L2 lexical items to the concepts are established over time, allowing for eventual native-like performance in both languages.
 In general, where the L1 and L2 share characteristics, “positive” transfer can be a helpful shortcut allowing learners to use their knowledge of their native language in order to perform in a native-like manner in the L2. But where the L1 and L2 diverge, “negative” transfer becomes evidence of the learner’s non-native status.
 They may, however, be slower to acquire language for spatial categories that are typologically rare (Gentner and Bowerman 2009).
 The categories reflected in the grouping of spatial scenes under shared linguistic labels are taken to be isomorphic to the underlying conceptual categories, an assumption supported by experimental work involving non-linguistic spatial categorization tasks (e.g., McDonough, Choi, and Mandler 2003; Athanasopoulos et al. 2015).
 ON is the focal category for this study, though uses of other English prepositions are also examined. These other prepositions correspond to Japanese spatial terms whose semantics partially align with the ON category. Time and space constraints prevented a full treatment of the semantics of all of these prepositions; the description of English spatial semantics here provides an detailed portrait of on, and additional information on the semantics of at is provided in the findings section as needed for purposes of differentiation.
 All pictures from the TRPS that are referred to in the text can be found at “L&C Field Manuals and Stimulus Materials,” Language and Cognition Department, Max Planck Institute for Psycholinguistics, http://fieldmanuals.mpi.nl/volumes/1992/topological-relations-picture-series/.
 The following abbreviations are used throughout the descriptions of English and Japanese to represent syntactic constituents: NP = noun phrase; P = preposition; PP = prepositional phrase; VP = verb phrase.
 Nonfinite verbs of the above-mentioned types, whose subject is the figure NP and whose PP complement contains the ground, are also included.
 The following description is based primarily on Kita (2006), the most extensive discussion of the grammar and semantics of Japanese topological relations that I found, with supplementation from other sources as noted. Kita’s discussion, as well as the representations of Japanese spatial semantics in Levinson and Wilkins (2006) more broadly and in Gentner and Bowerman (2009), does not account for mappings from certain Japanese spatial terms to certain ON relations identified by a native speaker with whom I consulted—for instance, from the spatial nominal naka, typically treated in the literature as denoting containment, to an English expression like on the train. Such mappings will be discussed briefly in the findings and conclusion sections.
 The following postpositions (particles) are used in the Japanese locative constructions to mark syntactic roles and cases: -wa = topic; -ni = dative; -ga = subject; -no = genitive (Kita 2006; Martin 1975). I follow Kita’s hyphenation scheme, which highlights the morphology. As for the translations of –ni, Kita gives only at and in, but Merriam-Webster’s Japanese Learner’s Dictionary (Merriam-Webster 1993), s.v. “ni,” gives definition (1a) as “at, in” and (1b) as “on, onto;” the Everyday Japanese Dictionary (McGraw-Hill Companies 2011) gives the definition as “at, to”; and Denji Jisho (http://jisho.org/word/に) gives sense (1) as “at, in, on, during.”
 Similar to English prepositions, ue is highly polysemous and has a range of extension beyond topological relations. For instance, it can be used with verbs of motion, as in sentences such as “The plane flew over the mountain”; it can refer to social superiority; and as a “postadnominal” (i.e., following a nominalized clause [Martin 1975]), it can have meanings such as ‘in addition to,’ ‘as a result of,’ ‘concerning,’ etc. See also Denshi Jisho, s.v. “ue,” http://jisho.org/words?jap=ue&eng=&dict=edict; Merriam–Webster’s Japanese-English Learner’s Dictionary, s.v. “ue.”
 Note that I treat over and above as synonyms, adopting a view that is limited to their function as prepositions in topological spatial descriptions (see Tyler and Evans  on their synonymy in this context and on the many other meanings of over).
 It appears that the use of ue is not obligatory whenever a specific region of the ground with which the figure is incontact can be distinguished. According to my native-speaker informant, in the realm of ON relations, only when this region is salient or the specific position of the figure is relevant is ue called for as opposed to –ni/-de alone.
 The prepositions given in these glosses should not be taken as direct translations from various Japanese postpositions; in all of these constructions, the ground NP is marked as dative with –ni (except in the case of the ring on the living finger, where the animacy of the ground forces the use of the nominative case).
 The corpus is distributed in CD-ROM format with an accompanying handbook (Granger et al. 2009) and can be purchased at www.i6doc.com/en/collections/cdicle/.
 The excluded contributor reported speaking English as a second language at home and having lived in an English- speaking country for 18 years. Thus, his English production is a priori likely to be more similar to that of a native English speaker than to that of an intermediate-level learner. Nevertheless, including this text in the sample would not substantively change the results reported below.
 Information about their geographic origins is not available from the ICLE. As reported in table 1, over half of the contributors were students at two universities in Tokyo, and many smaller groups of contributors were at other universities in Tokyo. It is unclear whether this has any bearing on the uniformity of their Japanese dialects, and it is important to note that Kita’s (2006) description of spatial language in Japanese is based only on the dialect of metropolitan Tokyo.
 The English proficiency of individual contributors to the ICLE was not measured. The compilers of the corpus recruited contributors only from among university undergraduates in English to roughly control for this variable. They conducted a post-hoc test of proficiency by sending a sample of each subcorpus to a professional rater for evaluation. While the contributors from some countries were rated as having advanced proficiency, the Japanese contributors were almost uniformly rated at the intermediate level (Granger et al. 2009, table 6).
 Checking the results of POS-tag-based searches in the ICLE is recommended by Granger et al. (2009, p. 16), since in the assembly of the corpus the results of the automated POS-tagging were not manually verified. The polysemy of prepositions and the ambiguity of some distinctions among free combinations, restricted collocations, and idioms does appear to have affected the accuracy of the POS-tagging for prepositions in the Japanese subcorpus in general. A number of occurrences of on in phrasal verbs or fixed idioms appeared in the search results for <on.II>, though by the logic of the CLAWS7 taxonomy these should have received the tags <.RP> and <CDIC>. However, most literal uses of on with a spatial meaning appear to have been classified appropriately with the basic preposition tag <.II>.
 Denji Jisho, s.v. “chikyuujou,” http://jisho.org/word/地球上; Merriam–Webster Japanese–English Learner’s Dictionary, s.v. “chikyuu”; Kenkyusha’s New Japanese-English Dictionary (Koine 1980), 5th ed., s.v. “chikyuu.”
 The latter is not discussed by Kita (2006), but the definition supplied by my informant is confirmed by Denji Jisho, s.v. “utsutteru,” http://jisho.org/word/映る.
 All 3 of these sentences apply to descriptions of people or objects “on” trains. There is no evidence in the ICLE essays from which they are drawn to suggest that the figures are on top of the roofs of these trains rather than within their interiors. Yet, as noted by Herskovits (1986, p. 144), in the case of “physical object[s] transported by a large vehicle … on is used, although containment is salient.” Thus, the lack of error in these examples shows mastery of a difficult case in English spatial semantics; examination of uses of in in the ICLE could reveal whether a higher proportion of learners fail in this case.
 The raw frequencies of <over.II> and <above.II> were, respectively, 127 and 11, for normalized frequencies per 100,000 words of 64.4 and 5.5; these were dispersed over 92 texts (25.2% of the whole sample) and 11 texts (3%), respectively. Thus, these two words are used relatively rarely in the ICLE as prepositions of any type, let alone as topological prepositions.
 For an example, see Aizu Tourist Information Centre, 2014, “Aizu Kiri Geta (Japanese sandals),” December 13, http://aizutouristinformationcentre.tumblr.com/post/105050035319/aizu-kiri-geta-japanese-sandals.
 The fact that this same error occurs in 5 different essays, by learners at 4 different universities, could indicate that the error has been propagated in Japan by some widely used textbook.
 Available at http://corpus.byu.edu/coca.
 Another factor that could affect the likelihood of transfer is the amount of time spent by the learners in an English- speaking country. However, as shown in table 1 above, this information is missing for more than half of the contributors in the sample, so analysis of this potential effect is forestalled.
 Most of Slobin’s work refrains from predicting any effects on non-verbal cognition, and the TFS hypothesis has not been widely applied to research in L2 acquisition. It therefore remains an open question whether TFS for a bilingual speaker takes place in the speaker’s L1, regardless of the language to be produced, or whether the speaker thinks in the L1 for speaking in the L1 and in the L2 for speaking in the L2. In an early work, Slobin (1996b, p. 89) postulated that TFS shaped by an L1 can cause non-native-like performance in an L2, prefiguring Ellis’s (2006b) description of associative learning: “Each native language has trained its speakers to pay different kinds of attention to events and experiences when talking about them. This training is carried out in childhood and is exceptionally resistant to restructuring in adult second-language acquisition.” While Slobin (2003, p. 159) later stated that “serious study of language in use points to pervasive effects of language on selective attention and memory for particular event characteristics,” he further noted, “More elusive have been clear demonstrations that these sorts of online attention may also have long-term and pervasive effects on mental representation and conceptual processes” (p.179). There are in fact some studies that apply the TFS hypothesis to L2 acquisition and find evidence that learners in some cases express themselves in their L2 using formulations congruent with their L1 (Choi and Lantolf 2008; González 2010), but this strain of the literature remains underdeveloped.
Athanasopoulous, Panos, et al. 2015. “Two Languages, Two Minds: Flexible Cognitive Processing Driven by Language of Operation.” Psychological Science. DOI: 10.1177/0956797614567509.
Biber, Douglas. 1990. “Methodological Issues Regarding Corpus-Based Analyses of Linguistic Variation.” Literary and Linguistic Computing 5:257–69.
Bowerman, Melissa, and Soonja Choi. 2001. “Shaping Meanings for Language: Universal and Language-Specific in the Acquisition of Spatial Semantic Categories.” Pp. 475–511 in Language Acquisition and Conceptual Development, ed. Melissa Bowerman and Stephen C. Levinson. Language Culture and Cognition no. 3. New York: Cambridge University Press.
———. 2003. “Space under Construction: Language-Specific Spatial Categorization in First- Language Acquisition.” Pp. 387–428 in Language in Mind: Advances in the Study of Language and Thought, ed. Susan Goldin-Meadow and Dedre Gentner. Cambridge, Mass: MIT Press.
Bowerman, Melissa, and Eric Pederson. 1992. “Topological Relations Picture Series.” In Space Stimuli Kit 1.2: November 1992, 51, ed. Stephen C. Levinson. Nijmegen: Max Planck Institute for Psycholinguistics.
Conklin, Kathy, and Norbert Schmitt. 2008. “Formulaic Sequences: Are They Processed More Quickly than Nonformulaic Language by Native and Nonnative Speakers?” Applied Linguistics 29 (1): 72–89.
Cosme, Christelle, and Gaëtanelle Gilquin. 2008. “Free and Bound Prepositions in a Contrastive Perspective: The Case of With and Avec.” Pp. 259–74 in Phraseology: An Interdisciplinary Perspective, ed. Sylviane Granger and Fanny Meunier. Amsterdam: John Benjamins.
Degani, Tamar, Alison Tseng, and Natasha Tokowicz. 2014. “Together or Apart: Learning of Translation-Ambiguous Words.” Bilingualism: Language and Cognition 17 (4): 749–65.
Dong, Yanping, Shichun Gui, and Brian MacWhinney. 2005. “Shared and Separate Meanings in the Bilingual Mental Lexicon.” Bilingualism: Language and Cognition 8:221–38.
Ellis, Nick C. 2006a. “Language Acquisition as Rational Contingency Learning.” Applied Linguistics 27 (1): 1–24.
———. 2006b. “Selective Attention and Transfer Phenomena in L2 Acquisition: Contingency, Cue Competition, Salience, Interference, Overshadowing, Blocking, and Perceptual Learning.” Applied Linguistics 27 (2): 164–94.
Garrod, Simon, Gillian Ferrier, and Siobhan Campbell. 1999. “In and On: Investigating the Functional Geometry of Spatial Prepositions.” Cognition 72 (2): 167–89.
Gentner, Dedre, and Melissa Bowerman. 2009. “Why Some Spatial Semantic Categories Are Harder to Learn than Others.” Pp. 465–80 in Crosslinguistic Approaches to the Psychology of Language: Research in the Tradition of Dan Isaac Slobin, ed. Jiansheng Guo, Elena Lieven, Nancy Budwig, Susan Ervin-Tripp, Keiko Nakamura, and Şeyda Özçalışkan. New York: Taylor & Francis.
Granger, Sylviane, Estelle Dagneaux, Fanny Meunier, and Magali Paquot, eds. 2009. International Corpus of Learner English, ver. 2 (handbook). Louvain: Presses universitaires de Louvain.
Gumperz, John, and Stephen Levinson. 1996. Introduction to Part I. Pp. 21–36 in Rethinking Linguistic Relativity. Studies in the Social and Cultural Foundation of Language 17. Cambridge: Cambridge University Press.
Herskovits, Annette. 1986. Language and Spatial Cognition: An Interdisciplinary Study of the Preposition in English. New York: Cambridge University Press.
Howarth, Peter. 1998. “Phraseology and Second Language Proficiency.” Applied Linguistics 19 (1): 24–44.
Jarvis, Scott. 2007. “Theoretical and Methodological Issues in the Investigation of Conceptual Transfer.” Vigo International Journal of Applied Linguistics, no. 4, 43–71.
Jarvis, Scott, and Aneta Pavlenko. 2008. Crosslinguistic Influence in Language and Cognition. London: Routledge.
Jiménez Catalán, Rosa María. 1996. “Frequency and Variability in Errors in the Use of English Prepositions.” Miscelánea: A Journal of English and American Studies 17:171–87.
Johnson, Jacqueline S., and Elissa L. Newport. 1989. “Critical Period Effects in Second Language Learning: The Influence of Maturational State on the Acquisition of English as a Second Language.” Cognitive Psychology 21:60–99.
Koine, Yoshio. 1980. Kenkyusha’s New English-Japanese Dictionary. Tokyo: Kenkyusha. Levinson, Stephen, Sérgio Meira, and The Language and Cognition Group. 2003. “’Natural Concepts’ in the Spatial Topological Domain—Adpositional Meanings in Crosslinguistic Perspective: An Exercise in Semantic Typology.” Language 79 (3): 485–516.
Levinson, Stephen C., and David P. Wilkins, eds. 2006. Grammars of Space: Explorations in Cognitive Diversity. New York: Cambridge University Press.
Mahmoodzadeh, Masoud. 2012. “A Cross-Linguistic Study of Prepositions in Persian and English: The Effect of Transfer.” Theory and Practice in Language Studies 2:734–40.
Martin, Samuel Elmo. 1975. A Reference Grammar of Japanese. New Haven: Yale University Press.
Merriam-Webster. 1993. Merriam-Webster’s Japanese-English Learner’s Dictionary. Springfield, Mass.: Merriam-Webster.
McDonough, Laraine, Soonja Choi, and Jean M. Mandler. 2003. “Understanding Spatial Relations: Flexible Infants, Lexical Adults.” Cognition 46:229–59.
McGraw-Hill Companies. 2011. Everyday Japanese Dictionary: English-Japanese, Japanese- English. New York: McGraw-Hill.
Nesselhauf, Nadja. 2005. Collocations in a Learner Corpus. Amsterdam: John Benjamins. Paquot, Magali, and Sylviane Granger. 2012. “Formulaic Language in Learner Corpora.” Annual Review of Applied Linguistics 32:130–49.
Pavlenko, Aneta. 2002. “Conceptual Change in Bilingual Memory: A NeoWhorfian Approach.” Pp. 69–94 in Advances in the Neurolinguistics of Bilingualism, ed. Franco Fabbro. Udine, Italy: Forum.
Reshöft, Nina, and Linn Gralla. 2013. “On the Use of Spatial Prepositions: Differences in L1 and L2 English.” Pp. 389–400 in Twenty Years of Learner Corpus Research: Looking Back, Moving Ahead, ed. Sylviane Granger, Gaëtanelle Gilquin, and Fanny Meunier. Louvain: Presses universitaires de Louvain.
Shinkawa, Susan H. 1979. “English ‘at,’ ‘in,’ ‘on,’ and ‘by’ Compared with Japanese de, ni, and o.” Pp. 160–77 in English and Japanese in Contrast, ed. Harvey M. Taylor. New York: Regents.
Slobin, Dan. 1996a. “From ‘Thought and Language’ to ‘Thinking for Speaking.’” Pp. 70–96 in Rethinking Linguistic Relativity, ed. John Gumperz and Stephen Levinson. Studies in the Social and Cultural Foundation of Language 17. Cambridge: Cambridge University Press.
———. 1996b. “Two Ways to Travel: Verbs of Motion in English and Spanish.” Pp. 195–219 in Grammatical Constructions: Their Form and Meaning, ed. Masayoshi Shibatani and Sandra Thompson. Oxford: Clarendon Press.
———. 2003. “Language and Thought Online: Cognitive Consequences of Linguistic Relativity.” Pp. 157–92 in Language in Mind: Advances in the Study of Language and Thought, ed. Susan Goldin-Meadow and Dedre Gentner. Cambridge, Mass.: MIT Press.
Talmy, Leonard. 1978. “Figure and Ground in Complex Sentences.” Pp. 625–49 in Universals of Human Language, vol. 4, ed. Joseph H. Greenberg et al. Stanford, Calif.: Stanford University Press.
Trenkic, Danijela, Jelena Mirkovic, and Gerry Altmann. 2014. “Real-Time Grammar Processing by Native and Non-native Speakers: Constructions Unique to the Second Language.” Bilingualism: Language and Cognition 17 (2): 237–57.
Tyler, Andrea, and Vyvyan Evans. 2003. The Semantics of English Prepositions: Spatial Scenes, Embodied Meaning and Cognition. New York: Cambridge University Press.
Yuan, Huei-Chun. 2014. “A Corpus-Based Study on the Influence of L1 on EFL Learners’ Use of Prepositions.” Theory and Practice in Language Studies 4:2513–21.
Zhang, Yuan, Norman Segalowitz, and Elizabeth Gatbonton. 2011. “Topological Spatial Representation across and within Languages: IN and ON in Mandarin Chinese and English.” Mental Lexicon 6 (3): 414–45.