Comprehension of idioms

Non-compositional models in general assume that idioms are stored like long words in memory.[3] For example, the entire phrase kick the bucket is represented by the meaning to die. The words kick and bucket separately do not contribute to the idiomatic meaning. Since individual words do not matter in an idiomatic expression, they are thought to be processed together as one entity during comprehension. In contrast, processing a literal expression requires the meaning of each word to be derived and then perceived in relation to the other words in the phrase. Non-compositional models include the literal first hypothesis, lexical representation hypothesis, and the direct access hypothesis.

The Literal First Hypothesis

• The literal first hypothesis suggests that an idiom is first processed literally upon encounter.[4] Idioms are often ambiguous since they may be processed in two ways: literally and idiomatically. For example, kick the bucket can be interpreted as to strike the pail with one’s foot. The literal first hypothesis proposes that, only after an appropriate literal meaning is not found, a mental idiom list is assessed to retrieve the idiomatic meaning. This hypothesis posits that there are separate storage for normal words and idioms, and that literal processing and figurative processing require different modes. Researchers claimed that this hypothesis was supported when people reported that they made an idiomatic interpretation first after they have read a set of idiom-containing sentences, but a literal interpretation first if they have seen a set of literal but ambiguous sentences.

The Lexical Representation Hypothesis

• Contrary to the previous model, the lexical representation hypothesis posits that idioms are stored along with normal words in memory.[5] Idioms are processed both literally and figuratively simultaneously. However, the figurative meaning (lexical representation) of the idiom is accessed first since the word string is stored and processed as one entry. Literal processing consumes more time since it requires the understanding of each word individually. Research conducted to support this hypothesis found that people recognized grammatical idioms as meaningful expressions more quickly than non-idiomatic but grammatical phrases when they were presented out of context.

The Direct Access Hypothesis

• The direct access hypothesis proposes that literal processing does not necessary occur when an idiom is encountered. It is possible that people completely bypass such mode when faced with a highly familiar idiom or have sufficient context to infer an idiomatic interpretation. In these cases, the idiomatic meaning is directly accessed.[6] This hypothesis is based on the findings that demonstrated that people processed idiomatic sentences quicker than literal ones, and tended to interpret idiomatic expressions figuratively even when they are used in a literal sense.

There are several criticisms for the non-compositional models. First, studies demonstrated that idiom expressions are not processed more slowly than literal expression. In fact, it is often the opposite, which is in contrary to the prediction of literal first hypothesis. Second, idioms have been found to be more than just “frozen phrases” or long words. For example, some idioms can be transformed to some extent and still be recognized and understood. For example, spill the beans can be used as “the beans were spilt by Mary”. This is possible because spill the bean can be mapped on the meaning “reveal the secret”, i.e. spill (reveal) and beans (secret). Such idiom shows that the internal structure of the word string matters during comprehension. The meaning of some idioms, like play with fire, can be also inferred from literal interpretation of their components (i.e. to do something dangerous). These findings reveal that idioms are not a homogeneous, distinct group and thus may not involve different processing strategies from those for literal expressions.[7]

Compositional models assume that idiom comprehension uses normal language processing.[3] When an idiomatic expression is encountered, it is processed incrementally like a normal expression. Components of an idiomatic word string contribute to a figurative meaning in either a literal or metaphorical way. Compositional models include the configurational hypothesis and the conceptual metaphor hypothesis.

Compositional models.

The Configurational Hypothesis

• The configurational hypothesis rose in response to the former hypotheses. It was based on a finding that demonstrated that people did not make an idiomatic interpretation once the first part of an idiom is encountered (against the prediction of the direct access hypothesis) or at the end of the idiom (against the prediction of the lexical representation hypothesis). Instead, the configurational hypothesis posits that a sufficient portion of an idiomatic word string must be processed literally before the idiom is identified.[8] The point where most people recognize the string as an idiom is the idiom key. Once the key is encountered, the words in the rest of the string will not be processed literally. The idiom key is the most important part of the idiom. If it is replaced by other words, the idiomatic meaning cannot be activated.

The Conceptual Metaphor Hypothesis

• The conceptual metaphor hypothesis proposes that metaphors are fundamental to human thought and they influence the comprehension of many aspects of language, including idioms.[9] An example of a conceptual metaphor is “love is a journey”. This metaphor is embedded in the idiom-containing sentences like their marriage is on the rocks and our relationship is at a cross-road. This hypothesis suggests that conceptual metaphors facilitate the understanding of idiomatic expressions. Individual words in the idiom can metaphorically contribute to its figurative meaning. In fact, research demonstrated that conceptual metaphors are activated during idiom comprehension. The words associated with the metaphor (e.g., journey) were more quickly identified as meaningful after the presentation of the idiom.

The prediction of the configurational hypothesis was not supported by research findings. Researchers found that even after a word string is recognized as an idiom, its literal meaning is still activated. Another criticism of the compositional models concerns the role of familiarity in idiom comprehension. As discovered by research on non-compositional models, idiomatic expressions are processed faster than non-idiomatic expressions. This is likely due to people’s familiarity of idioms. This suggests that when highly familiar idioms are encountered, people may not need to engage in literal processing or utilize conceptual metaphor to infer their meanings.[7]

Idiom familiarity is typically defined as how frequently an idiom is encountered in a language community. Subjective ratings of idiom familiarity are usually obtained from members of the language community. An example of a more familiar English idiom is pain in the neck, while a less familiar idiom is paddle his own canoe. Research demonstrated that familiar idioms are processed quicker and more accurately than unfamiliar ones.[11] As the dual idiom representation model suggests, highly familiar idioms may lead to direct memory retrieval. Processing unfamiliar idioms, on the other hand, requires contextual information and commonsense knowledge.

Idiom transparency can refer to the “literalness” of an idiom; how easy it is to understand an idiom based on the words it contains. Idioms can be sorted depending on their degree of transparency. Three categories of idioms have been identified: decomposable, abnormally decomposable, and nondecomposable.[6] Decomposable idioms are composed of words that literally contribute to their overall figurative meaning, e.g. pop the question. Pop refers to sudden and the question refers to marriage proposal. Abnormally decomposable idioms contain words that are associated with the overall figurative meaning of the idiom but in a metaphorical way, e.g. spill the beans. Spill mapping on to reveal and beans metaphorically representing secret. Nondecomposable idioms are made of words that do not reflect their idiomatic meaning, e.g. kick the bucket. People are found to respond to both types of decomposable idioms faster than nondecomposable ones.

Research in children highlighted the important effects of context on idiom comprehension. It was found that children understand idiomatic expression more accurately when they are shown in informative contexts than when they are presented in isolation.[12] When they are encountered out of context, children tended to interpret idioms literally. The ability to use contextual information in language processing has also been found to influence children’s performance in idiom comprehension. Adults, however, are more affected by the familiarity of the idiom.

Since aphasic patients often preserve their ability to use familiar phrases (i.e. idioms and proverbs), researchers hypothesized that they may be stored and processed in different brain regions than novel phrases. It was found that patients who experienced right-brain damage (RBD) showed more impairment in idiom comprehension than patients with left-brain damage (LBD). RBD patients however performed better in comprehending novel sentence compared to patients with LBD.[13] This suggests that the right hemisphere is involved in the comprehension of idioms. However, later research has suggested that RBD patients’ difficulties in idiom comprehension, as demonstrated by sentence-to-picture matching task, may be due to deficits in their visuospatial abilities, rather than impairment in linguistic processing. The right hemisphere shows dominance in visuospatial processing.[14]

Gray's Neuroanatomy.

Other researchers have hypothesized that comprehension of nondecomposable idioms mainly involved left hemispheric activity. When healthy adults were studied, left temporal rTMS was found to influence the reaction time and accuracy of both idiom comprehension and literal sentence processing. In particular, it increased processing time and reduced accuracy in understanding. When rTMS was applied to the left frontal, right frontal and right temporal, there were no effects on idiom comprehension.[15] Recent research with healthy adults using fMRI also found that, when processing idioms that were not literally interpretable (i.e., raining cats and dogs), the Broca’s area in the left prefronto-temporal network was activated. However, when literally interpretable idioms (i.e., break the ice) were processed, the left medial frontal gyrus (MFQ) and superior frontal gyrus (SFG) were activated.[16]

Prefrontal cortex (PFC) is suggested to be important for idiom comprehension. It may play a role in selecting the appropriate interpretation and suppressing the incorrect ones when an idiomatic expression is encountered. Research using fMRI indicated that the left temporal cortex, left superior medial frontal gyrus, and left inferior frontal gyrus were activated when idiomatic phrases were presented.[17] In addition, the right superior and middle temporal gyri, the right temporal pole, and the right inferior frontal gyrus were activated as well. In contrast, the left inferior parietal lobe and the right supramarginal gyrus were activated when literal sentences were presented.