下载掌阅APP，畅读海量书库
立即打开
Previous research focuses on three main questions concerning linguistic synaesthesia，including: （1）what transfer tendencies does linguistic synaesthesia show；（2）what are the underlying mechanisms for synaesthetic mappings；and （3）is the nature of linguistic synaesthesia neurological （Ronga et al.2012），metaphorical （Strik Lievers 2017），or literal （Winter 2019b）.
Existing studies on linguistic synaesthesia are mostly based on Indo-European languages and focus on the transfer directions of lexical items between sensory modalities. These studies have all demonstrated that linguistic synaesthesia follows directionality tendencies （Ullmann 1957；Williams 1976；among others）. Such tendencies are described in two different transfer models in the literature.
The first model originated from research on linguistic synaesthesia in poetic languages by Ullmann （1945，1957，1966 [1963]）. Based on around 2，000 synaesthetic examples collected from English，French，and Hungarian poems from the nineteenth century，Ullmann （1957）found that over 80% of the synaesthetic expressions conformed to the directional pattern described by the model in Figure 1.
Figure 1 A linear model for linguistic synaesthesia （adapted from Ullmann 1957）
Shen （1997）confirmed this linear model for Hebrew poetry. In addition，Shen （1997），Shen and Cohen （1998），Shen and Eisenman （2008），and Shen and Gadir （2009）demonstrated the validity of the model with various experimental studies on linguistic synaesthesia in non-poetic Hebrew and English，which involved tasks such as interpretation generation，recall，naturalness judgments，and so forth. For example，they have showed that synaesthetic expressions that conform to the directional tendency （e.g.， stinking paleness ）are easier to recall and are judged as more natural than expressions that violate the tendency （e.g.， pale stink ）. Shen and Gil （2008）conducted both corpus-based and experimental studies on linguistic synaesthesia in Indonesian，showing that the directional tendency described in Figure 1 also holds. However，all these studies on linguistic synaesthesia in non-Indo-European languages by Shen’s team are based on a small data sample，in terms of both the corpus data for generalizing directional tendencies and the testing stimuli utilized in the cognitive experiments. For instance，Shen （1997）utilized 130 synaesthetic instances of poetic Hebrew，and Shen and Gil （2008）employed 125 synaesthetic examples of non-poetic Indonesian to verify the transfer patterns of linguistic synaesthesia in each language. With respect to the experimental study by Shen and Gadir （2009），there are only two testing stimuli for each combination between any two of five senses.
Strik Lievers （2015）employed a larger data sample extracted automatically from the Wacky corpora （Baroni et al.2009）through a computational approach to test the linear transfer model for linguistic synaesthesia in non-poetic English and Italian. The study verified the model in Figure 1 for both English and Italian synaesthesia. Furthermore，Strik Lievers （2015：69）claimed that the directional tendencies of linguistic synaesthesia should be interpreted as“frequency”rather than as absolute constraints. The claim entails that linguistic synaesthesia shows directional preferences，where synaesthetic mappings in one direction occur more frequently than those in the reverse direction，rather than stating that the mappings in the reverse direction cannot be found in real languages. However，Winter （2019a）pointed out a problem in Strik Lievers’ （2015）corpus-based study with respect to methodology. That is，Strik Lievers （2015）did not take into consideration the distinction between types and tokens of synaesthetic transfers，both of which should be considered to determine the mapping directionality for linguistic synaesthesia （Winter 2019a）.
Williams （1976）proposed a second linguistic synaesthesia transfer model，as shown in Figure 2. By referring to the citation dates in the Oxford English Dictionary （OED）and Middle English Dictionary （MED），Williams generalized the model through focusing on the adjectives used for more than one sensory domain in the history of English. Williams assumed that English sensory adjectives would follow this model when they changed their meanings among sensory modalities，and that if synaesthetic transfers violated the pattern，the resultant sensory meanings would become obsolete in the language. For instance，the study illustrated that the tactile meaning of the adjective eager was not retained in Standard English，although the adjective had a tactile meaning as the result of a transfer from taste to touch in the history of English （see Williams 1976：476）. Furthermore，Williams suggested that the hierarchy described in Figure 2 should be applicable to any language，given the universality of the biological nature of human sensory modalities.
Figure 2 A transfer hierarchy for linguistic synaesthesia （Williams 1976：463）
Lehrer’s （1978）synchronic study on English sensory adjectives corroborated this model，but challenged the exclusive reliance on citations in dictionaries by Williams （1976），in consideration of the lack of written materials for earlier periods of English. She added that some unpredicted transfers by the model in Figure 2 could still be used in English，such as the dimension item fat （belonging to vision）employed for taste （Lehrer 1978：120）.
The two transfer models assign an identical directional rank to touch，taste，and smell，as seen in Figures 1 and 2. The main difference between the two models is the relationship between vision and hearing: while hearing precedes vision in Figure 1，both directions between hearing and vision are possible in Figure 2. The discrepancy is not，strictly speaking，contradictory，as Williams’（1976）model in Figure 2 classified vision into two different sub-domains: dimension and color （Zhao et al.2018b）. Since both are models of directional tendencies that do not claim to be strict rules，the same set of data could be consistent in both models. In addition，Winter （2019a）has observed that Williams’（1976）hierarchy of English synaesthesia is compatible with but stronger than Ullmann’s （1957）linear model （cf. Figure 1），as the former makes more falsifiable predictions and constraints for linguistic synaesthesia. In sum，studies on linguistic synaesthesia in Indo-European languages such as English and Italian，and non-Indo-European languages including Hebrew and Indonesian，have confirmed that different languages can share tendencies of directionality in synaesthetic transfers.
There is a debate on the mechanisms underlying linguistic synaesthesia. One theoretical model was proposed in the framework of Cognitive Linguistics，which generally treats linguistic synaesthesia as a specific type of metaphor （Geeraerts 2010）. This explanatory model claims that linguistic synaesthesia，similar to other types of metaphor，is also grounded in how our bodies experience the world，i.e.，the perceived similarity of intensity and subjective evaluation （Shen 1997；Yu 2003；Popova 2005；among others）. In this model，linguistic synaesthesia would follow the tendency of mapping from more embodied domains to less embodied ones，like other types of metaphor （see Lakoff and Johnson 1980；Johnson 1987）.
In addition，as the source domain and the target domain of linguistic synaesthesia are both bodily and concrete sensory modalities，Shen （1997）argued that additional features should be introduced to differentiate the degrees of embodiment of different sensory modalities. The study has pointed out that touch and taste necessarily involve physical contact between sensory organs and perceived objects，while the other three modalities （i.e.，vision，hearing，and smell）do not require such physical contact. Therefore，touch and taste are claimed to be more embodied than vision，hearing，and smell in Shen’s work.
Shen （1997）has also suggested that tactile perceptions，unlike other senses，do not require a specialized sensory organ，but instead have receptors all over the body. Hence，he proposed a rank of embodiment for five sensory modalities: touch is the most embodied sense，and taste follows，while vision，hearing，and smell are less embodied. Based on the ranking of the embodiment of five senses，Shen has contended that the directions of linguistic synaesthesia described in Figure 1 （cf.p.3）are constrained by the embodiment mechanisms，i.e.，mapping from the more embodied to the less embodied.
Similarly，Popova （2005：416）suggested that embodiment is realized“at its strongest in touch”，since scalarity and subjective evaluation of sensory stimuli dominate the perceptual experiences of touch （including taste），but not those of vision and hearing. Therefore，she assumed that the conceptualization of perceptual properties in vision and hearing in terms of concepts from touch and taste，illustrated the move from the more embodied to the less embodied.
The embodiment account for linguistic synaesthesia，however，has been criticized for its disregard of neural associations in the brain and of earlier occurrences of linguistic synaesthesia in the speech of young children compared to other kinds of metaphor. On the contrary，when we take the neural associations and the earlier occurrences of linguistic synaesthetic expressions into consideration，they might indicate that biological factors are at work （see Gardner 1974；Keil 1986；Marks et al.1987；Seitz 1997；Seitz 2005）. For instance，Ramachandran and Hubbard （2001：18）have proposed that linguistic synaesthesia is the same as neurological synaesthesia，as both are shaped by“anatomical constraints”that permit“certain types of cross-activation，but not others”. Seitz （2005：90）has pointed out that linguistic synaesthesia is an“inborn metaphorical association”that is“pre-wired”in brains.
Linguistic studies，such as Williams （1976）and Rakova （2003），also followed the biological association account for linguistic synaesthesia. For example，Rakova （2003）has suggested that the tactile meaning and the gustatory meaning of the English adjective hot are associated because of the same neural pain-detecting mechanism （i.e.，VR1，Caterina et al.1997），rather than through metaphorical mappings. In addition，Rakova （2003：64）has assumed that neurological synaesthesia is a“strong synaesthesia”and linguistic synaesthesia is a“weak synaesthesia”，both of which are analogous with respect to the innate neural associations between senses.
The biological association account of linguistic synaesthesia has also been challenged. For instance，Day （1996）compared types of neurological synaesthesia documented in Cytowic （2002 [1989]）with linguistic synaesthesia collected from English and German novels. The research found that the two kinds of synaesthetic phenomena did not share the same patterns. For instance，colored sounds （i.e.，the association between vision and hearing）were found to be the most common for neurological synaesthesia，while tactile sounds （i.e.，the association between touch and hearing）were the most frequent for linguistic synaesthesia （Day 1996）. Another criticism to the biological association account for linguistic synaesthesia is that the theory does not provide any explicit explanation for the directional tendency of linguistic synaesthesia （Popova 2005）.
Zhao et al.（2018b）tested the two rival theoretical models of mechanisms underlying linguistic synaesthesia，based on their different predictions of synaesthetic tendencies in languages. The study attempted to determine whether the directionality of mappings from the more embodied to the less embodied for linguistic synaesthesia predicted by the embodiment account，or the cross-linguistic universality of transfer patterns suggested by the biological association account，can be supported by real linguistic data. Based on plentiful data from the Sinica corpus （Chen et al.1996），the study found that distributions and tendencies of synaesthetic uses of gustatory adjectives in both Mandarin and English could not be explained or predicted by either single theory. Rather，linguistic synaesthesia can be accounted for by the combination of the two theories （Zhao et al.2018b）.
Linguistic synaesthesia，according to various studies，is a specific type of metaphor，and this view has become the“default”opinion in the literature （Strik Lievers 2017：87）. For instance，Osgood et al.（1978 [1957]）and Geeraerts （2010）suggested that linguistic synaesthesia shared a similar nature with other types of metaphor in terms of describing one concept through another concept based on similar features. Apart from the rhetorical framework for metaphors，the CMT-based framework （Lakoff and Johnson 1980；Lakoff and Johnson 1999）has also recognized the metaphorical nature of linguistic synaesthesia，which manifests itself as mappings of concepts from more embodied domains to less embodied ones （e.g.，Shen 1997；Yu 2003；Popova 2005）. Strik Lievers （2017）also argued for linguistic synaesthesia as a particular type of metaphor，as there are conceptual conflicts between sensory concepts from the different modalities involved. However，as pointed out by Rakova （2003）and Peng and Bai （2008），linguistic synaesthesia is not as the same as canonical metaphors that generally map from bodily domains to non-bodily domains （Steen 1999；Gibbs 2011）.
Following CMT，metaphor is not only a matter of words，but also a conceptual mechanism structuring human thoughts and actions （Lakoff and Johnson 1980；Johnson 1987）. Gibbs （2011）suggested that systematic metaphorical expressions using the concrete to describe the more abstract in different languages，such as LOVE IS A JOURNEY and ARGUMENT IS WAR，are crucial supporting evidence for metaphors not only in languages but also in conceptual systems. Nevertheless，linguistic synaesthesia has source domains and target domains that are both embodied and concrete sensory modalities. Thus，linguistic synaesthesia is different from canonical metaphors whose source domains exhibit an evident contrast with target domains concerning embodiment and concreteness.
Linguistic synaesthesia differs from canonical metaphors with respect to directional tendencies as well. Strik Lievers’（2015）corpus-based study has attested that linguistic synaesthesia does not follow absolute mapping directionality from the more embodied to the less embodied，as transfers violating this directional tendency can also be found in the language. For example，although transfers from touch to hearing occupy 23.2% of all English synaesthetic examples collected by Strik Lievers，there are still 0.2% of the total examples found to exhibit the transfer direction from hearing to touch （Strik Lievers 2015：80）. This pattern，however，has not been reported for typical metaphorical expressions mapping from the concrete to the abstract （see Lakoff and Johnson 1980；Sweetser 1990；Lakoff and Johnson 1999；among others）.
Theoretical explanations for mechanisms underlying linguistic synaesthesia and metaphors are also not identical. For instance，Rakova （2003：43）assumed that linguistic synaesthesia could not be explained by theories of metaphor，since both source and target domains of linguistic synaesthesia are bodily and concrete perceptions without evident contrasts on“experiential primacy”and“conceptual primacy”. Based on the corpus distribution of synaesthetic data，Zhao et al.（2018b）and Zhao and Huang （2018）have found that tendencies of linguistic synaesthesia cannot be accounted for or predicted completely by the embodiment mechanism，which has been widely recognized to underlie typical metaphors conceptualizing non-bodily experiences in terms of bodily experiences （Lakoff and Johnson 1980；Sweetser 1990；Gibbs 2011）.
Differences between linguistic synaesthesia and canonical metaphors can also be observed concerning research methodologies adopted. Studies on metaphorical expressions mainly begin with concepts expressed by lexical items （Steen 1999），based on which conceptual domains and mapping directions are determined subsequently （e.g.，Clausner and Croft 1999；Paradis 2001；Grady 2005；Lien 2005；Ou 2014；among many others）. Take the expression my mind just isn’t operating today for example （see Lakoff and Johnson 1980：27）. The contextual meaning of operating expresses a concept related to the mind. However，the lexical item operating has a literal meaning concerned with machines. Thus， operating can then be analyzed to illustrate the mapping from the conceptual domain of MACHINE to the conceptual domain of MIND in the example. Nevertheless，research on linguistic synaesthesia generally begins with the determination of sensory domains. Transfers of sensory items involving more than one sensory domain are then established （e.g.，Ullmann 1957；Williams 1976；Shen 1997；Strik Lievers 2015；among many others）. For instance，Williams （1976：476）first distinguished six sensory domains，and then determined that the English adjective sweet had the original sensory domain of taste. As the adjective can have an auditory use in contemporary English，such as sweet voice ，the study generalized the transfer direction from the sensory domain of taste to the sensory domain of hearing. Thus，unlike studies on metaphors that generally rely on a concept-based approach，studies on linguistic synaesthesia mainly employ a domain-based approach. It is intriguing that some canonical metaphor research，such as Ahrens （2002）and Chung et al.（2013），has also adopted the domain-based，rather than the concept-based，approach to examine metaphors，i.e.，to investigate the interaction between established ontological knowledge and conceptual mappings in metaphors.
In short，linguistic synaesthesia manifesting itself as a mapping from one bodily experience to another bodily experience exhibits distinct characteristics from canonical metaphors. These distinct characteristics include:（1）the contrast of embodiment between source domains and target domains；（2）directional tendencies of mappings；（3）theoretical explanations；and （4）research methods adopted.
Apart from the differences between linguistic synaesthesia and canonical metaphors，the metaphorical claim about linguistic synaesthesia has been challenged by the growing body of research on neurological synaesthesia. For instance，Ramachandran and Hubbard （2001）and Ronga et al.（2012）hypothesized a neurological basis for linguistic synaesthesia，based on the common origins and mechanisms between neurological synaesthesia and linguistic synaesthesia. Ronga et al.（2012）employed a corpus-based approach to show that linguistic synaesthesia has parallels with neurological synaesthesia concerning the patterns of sensory associations. Asona and Yokosawa （2012）and Hung et al.（2014）found that similar patterns could be observed in neurological synaesthesia among people speaking different languages from different cultures. This finding could provide a support for the similarity between linguistic synaesthesia and neurological synaesthesia，as Williams （1976）has suggested that transfer patterns of linguistic synaesthesia are cross-linguistically universal.
Winter （2019a，2019b）has claimed that linguistic synaesthesia is neither metaphorical nor neurological in nature，but instead it is literal，as sensory words have“highly multisensory or supramodal meanings”in language （Winter，2019a：107）. Take the English adjective sweet for example. Sweet fragrance was considered the literal usage of sweet，as the adjective is supramodal involving both taste and smell （Winter 2019b：125）. The assumption is similar to Rakova’s （2003）view，who suggested no distinction between literal and metaphorical meanings of sensory adjectives with no perceptual and conceptual priorities on any sensory modality. However，as pointed out by Popova （2005），this literal claim on the nature of linguistic synaesthesia would be inconsistent with the directional patterns of transfers in linguistic synaesthesia.
Therefore，the nature of linguistic synaesthesia has not been decided. It is still arguable whether linguistic synaesthesia can be properly analyzed as a specific type of metaphor，given the differences between linguistic synaesthesia and canonical metaphors and alternative assumptions on the nature of linguistic synaesthesia.