Conundrum 9:  The Whorfian hypothesis Summary

Introduction

The strong version of the Whorfian hypothesis is that language determines thought (i.e., language is the basis of thought). The weak version of the Sapir-Whorf hypothesis is that language influences thought. Linguistic representations can represent any type of information regardless of the sensory modality of the information (e.g., visual, auditory, thoughts, feelings) – that is why it is called an amodal representation.

Perception and language

Color perception

Colors provide objective visual information about our environment and humans can distinguish thousands of colors that vary in hue, saturation, and brightness. Color information provides us with very useful information about our environment that is utilized by higher-order cognition functions (e.g., object recognition, affect discrimination). The sensory processing of color involves three different types of cones in the retinas of the eyes while further color processing continues to multiple regions of the visual cortex (V1, V2, V4) and the inferior temporal cortex. Multiple regions of the brain are involved but cases of cerebral achromatopsia (colorblindness not related to missing cones) would suggest that some regions are more important for color perception. Much of color perception is performed automatically and unconsciously, such as color constancy that maintains the colors we perceive in our direct environment do not change even though their illumination may change dramatically (e.g., moving outside from inside).

However, color perception is not completely objective. Color perception is clearly subjective as shown by color illusions like the #TheDress illusion and can include bottom-up and top-down influences on the colors we perceive. Schlaffke et al (2015) tested individuals who differed in their perception of #TheDress illusion and found that regions of the brain associated with higher level cognitive functions showed increased brain activity in individuals who perceived one version of the illusion. Suggesting that higher-order cognitive processes can impact color perception. Language could be one of those processes. We primarily use only a few linguistic categories to distinguish different colors (e.g., red, blue), and the number of linguistic categories can vary markedly from one culture to another. The accurate use of these linguistic terms is one of the more challenging linguistic tasks of developing children with children as old as 4-5 years of age before they use these terms in spoken speech consistently and accurately. Even though research with infants suggests that they are already using these color categories before they have mastered the linguistic terms.

Rosch reported landmark research in the 1970s that tribes from Irian Java who only used two words for different colors could perform just as well as English speakers on a variety of color-related tasks (e.g., color perception, color memory). Her results suggested that there is no Whorfian effect, and that color perceptual processing is a universal mechanism found in all brains regardless of culture or language spoken. This was the dominant view for many years, but that view has changed as more sophisticated research methods (including the conducting of cross-cultural research) were developed.

Recent research by Roberson, Davidoff and colleagues (2000) used better mapping of linguistic boundaries of colors to test cross-category and within-category color memory judgments for English speakers (8 color categories) and Berinmo speakers (5 color categories) where category boundaries do not often overlap. Recognition memory performance was better for color chips if participants were tested with color chip pairs that came from different color categories (cross-categories) defined by their language rather than belonging to the same linguistic color category (within-category). These results support Whorf’s hypothesis.

Additional research has found further support for Whorf’s hypothesis using different language comparisons, although a review of this research by Kay and Reigner (2006) found that color processing across different languages may still have a universal brain mechanism that relates to the perception of focal (basic) colors regardless of language spoken. These small regions of colors were consistently found to fall within a linguistic category and not across two linguistic categories regardless of the language studied.

Spatial processing

Dehaene and colleagues (2006) tested an Amazon tribe on a variety of basic geometric and spatial concepts. These tribe members do not have words for such spatial concepts in their language. On some problems the Amazon tribe members did quite well, but for other spatial problems, they were performing at chance level. The researchers concluded that some form of core geometric knowledge could be a universal constituent of the human mind (i.e., not dependent on language spoken), but that language could still influence spatial processing (support for Whorf’s hypothesis).

Time processing

Boroditsky (2001) compared the effects of spatial metaphors on processing temporal information. English uses ‘horizontal’ (one-dimensional, directional) metaphors (e.g., getting ahead of schedule/falling behind schedule) for time whereas Mandarin speakers use ‘horizontal’ and ‘vertical ‘metaphors for time. English and Mandarin speaking participants were faster at answering temporal questions if primed with a horizontal prime question, but only Mandarin speakers showed a vertical priming effect for the control condition. These result support Whorf’s hypothesis that language can affect low-level cognitive processes.

Spatial-temporal processing

Finbeiner et al (2002) compared English, Japanese, and Spanish speakers on memory for moving objects that moved in different ways (manner) and followed different spatial paths of motion. Participants were shown an object moving in a specified manner and path using virtual reality. Participants were then shown a different display of the same object that either moved in the same manner (but not same path) or moved along same path (but not in same manner).  The researchers found that the English speakers (who possess many verbs for manner of motion, e.g., trot, scamper…) were more likely to view the manner alternative to be more like the test display than did the other language speakers (who do not possess many manner verbs). These results suggest that English speakers may perceive motion differently because of the language they use, and this finding would support Whorf’s hypothesis.

Language development and language loss

Language development

Does perceptual processing change as language changes, such as during the acquisition of color terms by children? Roberson et al (2004) examined color perception with English speaking and Otjihimba speaking children from the ages of 3-4 until they were 6-8 years of age (formative years of linguistic acquisition of color terms). They found that when the children were youngest there were no differences between the two language groups in how they performed a color memory task. Their performance on this color task was influenced by the perceptual distance between the stimuli being tested (i.e., their objective color properties). However, as the children got older and learned the names for various colors, color naming effects were now seen in the patterns of responses they provided for color-related tasks. Differences between the two groups now matched their languages spoken. These results support Whorf’s hypothesis. But these researchers did not find any support for the universal color processing mechanism suggested by Kay and Reigner (2006), because the focal colors were not the first color names learned by their child participants and there was no performance advantage when the color memory stimuli came from focal colors. Those findings would have been expected if there were universal color processing mechanisms that are independent of language spoken,

Kwok and colleagues (2011) trained adult participants to associate novel color names with specific colors. They showed gray matter changes in regions of the brain (visual cortex) associated with color processing because of this training. These findings suggest that learning new color names produced corresponding changes to the neural systems involved in processing color information, and these findings provide very strong support for Whorf’s hypothesis (if replicated).

Goldin-Meadows and colleagues have examined developmental trends in the use of gestures in communication. In one study, she found that young children who had yet to understand the conservation principle (quantity doesn’t change when it’s altered in some way – such as different glass shapes holding same amount of liquid), provided different gestures when giving their wrong answers. When water from one tall skinny glass was poured into a low wide container, these children reported that the amount of water contained in the lower/wider glass was less. This error reflects their cognitive development because they are still processing object size in one dimension (i.e., height). But some children used a different gesture that relates to the width of the glass rather than to the height of the glass when making this error. The children who produced this gesture-speech mismatch were younger when they understood the conservation principle that corrects for this error (i.e., consider two dimensions). Goldin-Meadows proposed that this gesture-mismatch suggests these children have access to some form of correct thinking, but don’t have the language to express these thoughts? This would be a problem for Whorf’s hypothesis.

Researchers (including Goldin-Meadows) have also examined the development of sign languages in children who did not receive any formal training in a language’s sign language equivalent. These sign languages are unique to these individuals, but the sign languages that emerge have many of the linguistic properties associated with spoken languages. How could these individuals develop a novel sign language without a language to represent their corresponding thoughts? This is a big problem for the Whorfian hypothesis.

Research with bilinguals who have immigrated to a new country (consecutive bilinguals) has shown that the language these bilinguals used to communicate at the time they experienced the event is more likely to be retrieved in that language from memory. Other research by Cox et al (2019) has shown that memories from an earlier language are more likely to be recalled in the language related to their prior experience with the language (e.g., born in Spanish speaking country) rather than their fluency in the language. These results suggest that autobiographical memories may be stored as linguistic representations (also supported by the language explanation for childhood amnesia in a previous conundrum) and this explanation would support the Whorfian hypothesis.

Language loss

Patients with aphasia have lost some (and possibly many) language functions due to brain damage. One major distinction between the types of aphasias documented was established in the 1800s with the identification of Wernicke’s aphasia (damage to temporal cortex and results in language comprehension problems) and Broca’s aphasia (damage to frontal cortex and results in language production problems). But distinguishing the different types of language loss found in patients with aphasia is very challenging, and many modern complicated diagnostic classification systems are currently in use (e.g., the 60 sub-tests of the PALPA). Language is a complex high-order cognitive function (as we will see again in the next conundrum) and aphasic patients have great difficulty describing their cognitive loss and providing verbal responses to cognitive tests. This makes it very challenging to make strong conclusions regarding the relationship between thought and language loss in this patient group.

Davidoff (2001) reported the case study of an aphasic patient who did not categorize facial emotions in terms of the standard basic linguistic categories for emotions that normal participants use (e.g., ‘fear’, ‘sadness’, ‘happiness’, ‘surprise’, ‘disgust’). The aphasia patient categorized the faces on how similar they looked perceptually. Does this mean that the aphasia patient cannot understand the emotions expressed by others because he has lost access to the linguistic labels for these emotions? If so, this would provide strong support for the Whorfian hypothesis.

Varley and Siegel (2000) examined an aphasic patient who could understand words but could not understand sentences (agrammatism aphasia), and yet, this patient showed an ability to perform reasoning tasks correctly. These researchers concluded that the patient could still perform cognitive reasoning without the linguistic tools needed to express the reasoning task, and this would be a problem for the Whorfian hypothesis.

Varley and colleagues (2005) tested agrammatism patients on mathematical tasks which we assume rely on linguistic translations of mathematical expressions to understand and solve. The aphasic patients did very well on these math tasks, and this finding suggests that access to linguistic representations is not necessary for the cognitive processing of such mathematical problems. These results are a problem for the Whorfian hypothesis.

Frontal dynamic aphasia patients can process words and sentences but have trouble constructing narratives and stories. Robinson et al (2006) tested a dynamic aphasia patient’s descriptions of events depicted in pictures (e.g., the cookie theft picture). The patient could describe the story but there were prolonged pauses during the spoken response. What was going on in the patient’s head during these pauses and what was causing these long pauses? Perhaps, she was still thinking but forced to take excessive time to find the correct verbal description to use. If that was the case, this finding would also be a problem for Whorfian hypothesis because there was thought without language.

Lazar and colleagues (2000) temporarily induced a form of Wernicke’s aphasia in an awake patient by anaesthetizing the area of the brain associated with Wernicke’s aphasia. The patient could not provide correct overt responses to linguistic tests (e.g., word meaning), even though he was aware of what he was thinking at the time and that he thought he had the correct answer in his head. Could this be another example of thoughts without access to language and another problem for Whorf’s hypothesis?

Language of thought

There is a long history of scholars who have proposed that language is not the basis of thought. Influential Russian researchers, Vgotsky and Luria, suggest that thought comes before inner speech and that inner speech comes before overt speech. Pinker (1994) provided several arguments why language could not be the basis for thought. Some alternatives to linguistic (symbolic, amodal) representations have been proposed.

Inner Speech

One alternative is inner speech (inner voice) if it was different from overt speech. American behaviorist researchers (e.g., John Watson) on the other hand suggest that inner speech is simply the same as overt speech. Research on inner speech is challenging. One method that has been used is articulatory suppression (repeatedly saying a word or phrase to yourself in your head) to block the phonological loop in working memory. Does blocking inner speech disrupt thinking as measured by a variety of cognitive tasks? Unfortunately, the results are mixed using this methodology.

Anecdotal evidence from aphasia patient case studies suggest that these patients still use inner speech without access to overt speech, and that their inner speech is correct even though their overt speech response is not. Perhaps, the aphasia patients described previously in the reasoning and mathematics studies were using inner speech to successfully perform these cognitive tasks and therefore still had access to linguistic representations.

Embodied cognition

Embodied cognition theorists propose that cognitive concepts are represented as mental simulations of perceptions and actions, and these simulations are used in cognitive processing.

Hauk and colleagues (2004) showed that when participants read action verbs (e.g, walk) brain activity results in the somatosensory and motor regions of the cortex associated with controlling the muscles/limbs (e.g., foot) associated with these actions. The participants are perhaps reexperiencing the sensory and motor aspects that are involved in the actions represented by these words. In a similar fashion, embodied cognition researchers would suggest that memories of events we have experienced are stored as the sensory and motor activity we experienced at the time (rather than linguistic descriptions of the event).

The participants in Zwann and colleagues (2004) study performed a size perception task with a second ball displayed after a first ball where the balls could be the same size or differ in size (second ball shape or be smaller or larger than first ball shape). While the participant was making this size perceptual judgment (same or different in size), auditory sentences were being simultaneously played in the background. Even though the participants were told to ignore these sentences, their size perception judgments were faster if the size change conveyed in the perceptual task (e.g., second ball was smaller as would be the case if the ball was moving away from the participant) was congruent with the background auditory sentence describing a similar motion (e.g., “You tossed a beachball over the sand towards the kids.”). The researchers assumed that the linguistic processing of the sentence automatically produced simultaneous activation in the same regions of the brain used for processing the visual information needed for processing congruent visual stimuli in the perceptual task. This activation facilitated the processing of the perceptual regions of the brain needed to make these size judgments.

In another embodied cognition study, Mathot et al (2017) measured pupil size as participants read or listened to words that differed in their illumination details (e.g., ‘night’ v ‘sun’). Pupil size varied (involuntarily) as a function of the illumination associated with words, suggesting that the visual system (even before the visual information has reached the retina) has automatically and involuntarily simulated the sensory-motor aspects of the linguistic representations associated with words.

Embodied cognition theory has been successful explaining concrete concepts, such as a ‘cat’, but has difficulty explaining abstract concepts, like ‘freedom’. Hybrid approaches involving both embodied representations and linguistic representations are now the more popular theoretical explanations for the language of thought. Whorf’s hypothesis is therefore still relevant to these hybrid explanations.