Why experts get it wrong

Why experts get it wrong: being knowledgeable about a subject implants false memories

 Sarah Knapton,   the telegraph 

http://www.telegraph.co.uk/science/2016/08/24/why-experts-get-it-wrong-being-knowledgeable-about-a-subject-imp/

The Homophone Meaning Generation Test

  

When people are asked to supply as many meanings as they can for a word presented orally and out of context, they begin with the most frequent meanings, and once these are exhausted, turn to other, less common, meanings.  This process requires a strategic search in the knowledge base and mental flexibility, which are part of executive functions.

The homophone meaning generation test – HMGT , developed in Hebrew by Prof. Gitit Kave and her colleagues,   tests this process.  Homophones are words which have different meanings that "sound the same".  For instance, the word "right" has the meaning of "a direction (opposed to "left")" and the meaning of "appropriate, suitable".  

The test in its Hebrew form is made of 24 homophones, each having between three and ten possible meanings.  Half of the words are also homographs (all their meanings are spelled  the same way).  The words are presented orally, one by one.  The person is asked to think about as many meanings as he can to each word.  There is no time limit.  When the person is not able to think about any more meanings, the next word is presented.  Each different meaning is assigned a point.  The score in the test is the number of meanings given to all presented words.  The production of meanings requires strategic search and mental flexibility, thus the test is supposed to measure executive functions. 

Prof. Kave and her colleagues performed a few experiments with this test, in children and adults, two of them will be presented here, 

In the adult study, the relation between the homophone test and shifting and clustering in fluency tests was examined.

What are shifting and clustering?

Phonemic fluency tests (in which the person is asked to name as many words beginning with a specific letter within one minute as he can) and semantic fluency tests (in which a person is asked to say as many words belonging to a specific category within one minute as he can) measure the ability to search and retrieve information from long term memory (thus they belong to the broad ability "Long Term Storage and Retrieval").

In both of these tests, the memory search includes two aspects:  clusters and switches.  During test performance people usually produce clusters of words that are related to each other semantically (for instance, "horse, cow, donkey, goat, sheep" is a cluster of "farm animals") or phonetically (for example, "buy, bug, bus" is a cluster of words beginning with "bu").  When the subcategory from which the words were retrieved is exhausted (the person cannot think about more farm animals or more words beginning with "bu"), the person shifts to another subcategory (for instance, he will transfer to "sea animals" or to "be").

Clusters reflect the semantic organization of the crystallized knowledge.  Shifts reflect executive functions (strategic search in the fund of lexical knowledge, monitoring, the initiation of response, flexibility, an ability to shift set).

The phonemic fluency test is presumed to require executive functions more than the semantic fluency test.  This is because when we retrieve information by content (like in the semantic fluency test) we perform a familiar act which fits the way our knowledge is organized.  But when we retrieve words beginning with a specific letter, we perform a new and unfamiliar task.  In order to optimally perform this task we have to come up with a search strategy, monitor the search process and act flexibly.  These are executive functions.

One hundred volunteers aged 18-35 and with an average of 13.8 years of  education participated in the study.  All participants were healthy native Hebrew speakers.  Each participant was tested with the homophone test and the semantic and phonemic fluency tests.

Since the homophone test was built as an executive functioning test, requiring directed and flexible memory search, the assumption was that the homophone test will be correlated with phonemic fluency more than with semantic fluency.  In fact, a significant and equal correlation was found between the homophone test and both fluency tests.

The authors offer 2 kinds of explanations for this finding:  it may be that the lexical knowledge component existing in the homophone test and the semantic fluency test caused the correlation between these tests to be higher than expected.  Another possibility is that both fluency tests require a similar executive search.  The relation between all three tests can be resulting from a common executive component, which is necessary for successful performance in these tasks.

When we look at the shifts and clusters data we get a finer picture:

Performance in the homophone test was related more to the number of switches or the number of clusters that a person performed in both fluency tests, than to the average size of the cluster in these tests.  This strengthens the hypothesis, that the relation between the three tests is stronger in the executive component than in the lexical semantic component.  While members in a sematic cluster (for example, "dog, cat, hare, gerbil") belong to a similar conceptual field, this is not the case with homophone representations.    Thus performance in the homophone test cannot be attributed to the spread of activation within a sub category in the semantic lexicon 

Participants in this study produced more meanings for non-homographic homophones than for words that are both homophones and homographs.  The score of the homographs was more strongly related to the switching component in both fluency tests than the score on the non homographs.  This means that finding different meanings for homophones that are also homographs requires more executive functions than finding different meanings for homophones that are not homographs. 

How do children perform in the homophone test?  A child cannot retrieve meanings that do not yet exist in his lexicon.  But sometimes a child will find it hard to retrieve meanings he is familiar with since his retrieval skills may not be flexible enough.  Children's retrieval skills are affected by the development of their vocabulary and the development of controlled search processes in their existing lexical knowledge. 

Children's vocabulary is developing constantly, as the child is exposed to reading and literature.  At the end of 2^nd grade, English speaking children's vocabulary contains 6000 word meanings.  Their vocabulary increases in the next few years at a rate of 1000 meanings per year.  In addition to the growing lexicon, knowledge about word meanings also grows.  The meanings of new words are gradually refined in a process that continues into adolescence.

Efficient  search strategies also develop throughout childhood and into adolescence.  These search strategies are one of the manifestations of executive functions. 

In the children's study, changes in word retrieval throughout childhood were assessed by four tasks:  a picture naming task, a phonemic fluency task, a semantic fluency task and the homophone test.  These tasks differ in the amount of flexibility needed to perform the test. 

In the picture naming task (in which the child sees a picture of an object and is asked to name it) there is no need for flexibility in lexical search.  Each picture corresponds to one word only.  When this word is found, the search is over.

Semantic and phonemic fluency tests require a more flexible search in the lexicon, using clustering and switching strategies as was discussed earlier.  There is a developmental improvement in the performance on these tests up to the age of 12 and beyond.  The fluency component that improves the most throughout childhood is the shifting component.  As was said before, this is an executive component which reflects strategic search, response initiation, monitoring and flexibility.

In order to perform the homophone test optimally, the word search has to continue much after the first word has been retrieved.  Unlike the fluency tests in which the search is performed on a limited set of the lexical knowledge, the homophone test requires searching and retrieving from the whole lexicon.  Thus it may require a more flexible search strategy than is required for the other tests.

Two hundred and seven children participated in the study, 20 from each age group in the age range of 8-17 and in grades 3^rd to 12^th.  The children were born in Israel and speak Hebrew as a first language.  They don't have learning disabilities, neurological or developmental problems and their socioeconomic status is average.

The children's scores on all four tests rose steadily between age 8 and age 17.  Picture naming ability rose the least (although the test being used was probably too easy in its Hebrew version, since the youngest children succeeded in 83% of its items).    Scores on the homophone test showed the steepest rise.  Thus, performance on the homophone test changed the most during development.

Like the adults, children produced less meanings for homophones that are also homographs (all their meanings are spelled the same) than for homophones that are not homographs.  The rate of development in performance on the homographs was identical to the rate of development in performance on non homographs. 

To summarize, this test assesses executive functions in a verbal task.  High scores on the homophone test indicate that the person has a rich lexical network and that he is able to shift between meanings flexibly.

How would this test be classified in CHC abilities?

I've recently written about the place of executive functions in CHC theory – "between" fluid ability and short term memory. 

  

The homophone test does not fit any of the definitions of the Fluid  narrow abilities (induction, deduction, quantitative reasoning).  In Hoelzle's study, other tests measuring executive functions are classified under fluid ability.  Hoelzle did not classify the homophone test or similar tests.  Due to the nature of this test and the classification of other executive function tests to fluid ability, perhaps we can classify this test also to fluid ability, at least tentatively.

The homophone test does not fit any of the definitions of narrow abilities within Short Term Memory.

The narrow ability "Ideational fluency" (in "Long Term Storage and Retrieval") is defined as the "ability to rapidly produce a series of ideas, words, or phrases related to a specific condition or object.".  Because of the speed factor, the homophone test cannot be classified here.

The narrow ability "Lexical knowledge" (in "Comprehension  - Knowledge") is "knowledge of the definitions of words and the concepts that underlie them".  The homophone test does require such knowledge and seems to fit here.

So it seems that in CHC terms the test can be considered to test Fluid ability and Comprehension Knowledge (Lexical Knowledge).

References:

KavÉ, G., Avraham, A., Kukulansky-Segal, D., & Herzberg, O. (2007). How does the homophone meaning generation test associate with the phonemic and semantic fluency tests? A quantitative and qualitative analysis. Journal of the International Neuropsychological Society, 13(03), 424-432.‏  

Kavé, G., Kukulansky-Segal, D., Avraham, A., Herzberg, O., & Landa, J. (2010). Searching for the right word: Performance on four word-retrieval tasks across childhood. Child Neuropsychology, 16(6), 549-563.‏

Hoelzle, J. B. (2008). Neuropsychological assessment and the Cattell-Horn-Carroll (CHC) cognitive abilities model. ProQuest.‏    PAGE 114   http://utdr.utoledo.edu/cgi/viewcontent.cgi?article=2213&context=theses-dissertations

The brain dictionary

Amazing 3 min. video and amazing research.

Where exactly are the words in your head? Scientists have created an interactive map showing which brain areas respond to hearing different words. The map reveals how language is spread throughout the cortex and across both hemispheres, showing groups of words clustered together by meaning. The beautiful interactive model allows us to explore the complex organisation of the enormous dictionaries in our heads.

Explore the brain model for yourself here: http://gallantlab.org/huth2016 

Read the paper here:  http://www.nature.com/doifinder/10.10...

Comprehension –knowledge: a primary broad ability or a product of the application of other cognitive abilities?

Comprehension knowledge is the breadth and depth of skills and knowledge that are valued by one's culture.  Comprehension knowledge includes the ability to understand spoken language, the breadth of a person's lexicon and general knowledge, and a person's awareness of grammatical aspects of language.

Is comprehension knowledge an independent ability or a product of the application of other cognitive abilities?  In other words, is it a cognitive ability, like fluid ability, short term memory and long term storage and retrieval or is it an area of achievement like reading, writing and math?

As will be discussed below, it's possible to conceptualize comprehension knowledge both as an ability and as an achievement.

Language impairment (LI) is an impairment in expressive and/or receptive language development  in the context of otherwise normal development (i.e., nonverbal IQ and self-help skills).   Language impairment interferes with activities of daily living and/or academic achievement.   LI children usually have poor comprehension knowledge, although LI will not always manifest in poor scores on intelligence subtest scores measuring comprehension knowledge.  The reason for that is that these tests do not assess all aspects of comprehension knowledge and especially not the child's command of grammar and syntax.  

A broad distinction can be drawn between two classes of LI model: those that regard the language difficulties as secondary to more general nonlinguistic deficits, and those that postulate a specifically linguistic deficit.

 The best known example of the first type of model is the rapid temporal processing (RTP) theory of Tallal and colleagues, which maintains that language learning is handicapped because of poor temporal resolution of perceptual systems.   The first evidence for the RTP theory came from a study where children were required to match the order of two tones.  When tones were rapid or brief, children with LI had problems in correctly identifying them, even though they were readily discriminable at slow presentation rates.  Children who have poor temporal resolution will chunk incoming speech in blocks of hundreds of milliseconds rather than tens of milliseconds, and this will affect speech perception and hence on aspects of language learning.

In CHC terms Tallal's theory can be conceptualized as poor auditory processing and maybe also  poor processing speed that underlie LI. 

Another theoretical account that stresses nonlinguistic temporal processing has been proposed by Miller, who showed that children with LI had slower reaction times than did control children matched on nonverbal IQ on a range of cognitive tasks, including some, such as mental rotation, that involved no language. Unlike the RTP theory, this account focuses on slowing of cognition rather than perception.

 A more specialized theory is the phonological short-term memory deficit account of LI by Gathercole & Baddeley. These authors noted that many children with LI are poor at repeating polysyllabic nonwords, a deficit that has been confirmed in many subsequent studies. This deficit has been interpreted as indicating a limitation in a phonological short-term memory system that is important for learning new vocabulary  and syntax.  

Ullman argues that our use of language depends upon two capacities: a mental lexicon of memorized words and a mental grammar of rules that underlie the composition of lexical forms into predictably structured larger words, phrases, and sentences.   On this view, the memorization and use of at least simple words depends upon an associative memory (the ability to learn links between pairs of stimuli, for example between the sequence of phonemes forming a word and its meaning.  Associative memory is a narrow ability within long term storage and retrieval).  The acquisition and use of grammatical rules depends upon procedural learning and memory (for instance, learning the rule/procedure of past tense as the addition of "ed" (walk – walked)).   Procedural memory enables us to learn many motor and cognitive "skills" and "habits" (e.g., from simple motor acts to skilled game playing).  LI, argues Ullman, is caused by poor associative and/or procedural memory.  LI is not a specifically linguistic disorder but is rather the consequence of an impaired system that will also affect learning of other procedural operations, such as motor skills.

To all this we can add Cattell's investment theory.  Cattell argued  that crystallized knowledge develops through the investment of fluid ability.  Babies are born with fluid ability only, which they use in their first encounters and experiences with the world.   Explicit and implicit memories formed in the baby's mind by these encounters and experiences gradually form his reservoir of crystallized knowledge.  From now on, the baby tackles new experiences equipped with both fluid ability and crystallized knowledge that he already acquired.  The more knowledge he acquires, the better is his ability to cope with situations he encounters, and the less fluid these situations become.

The theories presented above enable us to consider comprehension knowledge as an area of achievement.  When a child performs poorly in an area of achievement (reading decoding, reading comprehension, writing and spelling, math computations or math reasoning) we look for the cause of his poor performance among the cognitive abilities (fluid ability, short term memory, long term storage and retrieval, processing speed, visual processing, auditory processing, comprehension knowledge).  If comprehension knowledge is also an achievement area, when a child has poor comprehension knowledge, we should look for the reason for that in one of these cognitive abilities:  fluid ability, short term memory, long term storage and retrieval, processing speed, auditory processing and perhaps also visual processing (I don't know a theory linking LI to poor visual processing).

But comprehension knowledge is different than other achievement areas.  Reading, writing and math are acquired mostly through formal instruction (there are aspects of arithmetic that are acquired naturally without instruction – like counting procedures).  Native language is acquired mostly informally – mostly spontaneously/implicitly.  Formal instruction enriches and develops native language.

The authors above argued that domain-general deficits in cognitive and perceptual systems are sufficient to account for LI. This position differs radically from linguistic accounts of LI, which maintain that humans have evolved specialized language learning mechanisms and that LI results when these fail to develop on the normal schedule  . A range of theories of this type for LI focus on the syntactic difficulties that are a core feature of many children with LI. Children with LI tend to have problems in using verb inflections that mark tense, so they might say “yesterday I walk to school” rather than “yesterday I walked to school.” Different linguistic accounts of the specific nature of such problems all maintain that the deficit is located in a domain-specific system that handles syntactic operations and is not a secondary consequence of a more general cognitive processing deficit.

Pennington, B. F., & Bishop, D. V. (2009). Relations Among Speech, Language, and Reading Disorders. Annu. Rev. Psychol, 60, 283-306.‏  http://www.du.edu/psychology/dnrl/Relationsamongspeechlanguageandreadingdisorders.pdf

Schneider, W. J., & McGrew, K. S. (2012). The Cattell-Horn-Carroll model of intelligence. Contemporary intellectual assessment: Theories, tests, and, (3rd), 99-144.‏

A slightly different view on fluid ability and its relations with working memory, comprehension knowledge and learning

 

Lohman, D. F., Lakin, J. M., Sternberg, R. J., & Kaufman, S. B. (2009).Reasoning and intelligence. Handbook of intelligence, 419-441.‏    

https://faculty.education.uiowa.edu/docs/dlohman/Reasoning-and-Intell_Lohman-Lakin-102709.pdf

This is a wonderful chapter about reasoning, which is part of the fluid ability.  Almost every sentence in this chapter is significant.  Here I'll focus on some of the ideas that were especially interesting to me.

 

Reasoning processes, comprehension knowledge and learning

. 

People who solve problems efficiently usually turn their attention, sometimes even unconsciously, to different aspects of the problem than people who solve problem less efficiently.  People who are efficient problem solvers know what to look for and what to ignore.  This results from more experience that they have with similar problems and from an ability to make good use of their past experience.

Good thinking about complex problems depends on knowledge.  Expertise is based on knowledge, and experts think about problems differently than novices. Good reasoning leads to better organization of knowledge in memory.  Every reasoning process we perform depends on the efficiency of the reasoning processes we performed in the past, and on the ways we stored the products of these processes in memory.  An increasingly sophisticated knowledge base supports increasingly sophisticated forms of reasoning. A more sophisticated knowledge base has richer, more abstract associative connections between concepts and more metacognitive knowledge that links strategies to goals.  In other words, the better our reasoning processes are, the more complex our comprehension knowledge, which is the product of these reasoning processes, will be.  

Reasoning abilities are apparent even in a vocabulary test.  Individual differences in a vocabulary test can result from the extent to which we use metacognitive processes when we learn words.  For instance, do we systematically examine alternative meanings of a word when we hear it in an unfamiliar context, or do we remain with a vague understanding of it? When  we infer that a word has additional meanings, and understand them with precision, we are able reorganize our comprehension knowledge in a way that will assist us in future learning.  Researchers argue, that the ability to derive word meanings out of the contexts in which they are heard may be the cause of the high correlation that usually exists between vocabulary and reasoning tests (and also between vocabulary and g).  A wide vocabulary enables the understanding and expression of a wider variety of ideas, and assists in learning new words and concepts.  Thus language functions as a tool for expression, refinement and acquisition of thought, and the modest vocabulary test requires reasoning processes and is a product of reasoning processes. 

  

Choosing what knowledge to apply to a new problem is a nontrivial source of reasoning complexity. Developmental psychologists have long known that children reveal much about the sophistication of their reasoning by how they classify or sort objects: on the basis of an arbitrary association, or by using perceptual characteristics, or, at the highest level, by using several different abstract concepts.  Therefore, deciding how best to describe the relationships among two or more concepts is a critical step in reasoning.  Poor reasoners often settle for a vague relationship or rule rather for a more exact one.  This could be because they terminate the search for a rule or relationship too quickly, or because they do not critically examine how well candidate rules or relationships describe the data, or because they simply do not see or know the rule.

It's possible to practice and train deductive thinking (applying known rules to specific cases), or at least, it's possible to improve performance in specific tasks that measure deductive thinking. This means that deductive reasoning tests can measure different abilities in examinees who have learned strategies for solving problems like those used on the test than for examinees who must invent a strategy on the spot.  For people who practiced, performance on such tests leans much more on the retrieval of the learned strategy out of the knowledge base.  For people who did not practice, these tests are much more fluid. 

Reasoning processes require ongoing activation of executive functions like monitoring, feedback, planning, adjusting existing strategies to the new situation and inventing new strategies, and the ability to learn from past attempts to solve the problem.  Reading comprehension, for example, requires updating our mental model of the text  with new information received during reading.  For instance, if we perceive one of the characters in our mental model as a bad person, and we read about a good deed this person did, we update our mental model of him.  Children with monitoring difficulties (executive function difficulties) find it hard to do that.  They may stick with the initial impression they formed of this character, and fail to update it with later incoming information. 

Naïve interpreters of reasoning tests think that reasoning ability influences achievement level (in other words, they see causal arrows running only from reasoning ability to achievement).  But both reasoning ability and achievement are products of leaning and experience. 

Understanding a story, inferring the meaning of an unknown word, recognizing patterns and rules in information, abstracting the given information in order to create rules or principles, applying mathematical concepts to solve a problem – in these and in many other ways, successful learning requires reasoning strategies.  The best way to develop reasoning is through challenging instruction that requires students to practice familiar reasoning strategies and to invent or to learn new strategies.  Thus, learning requires reasoning, but the learning process itself develops reasoning.

In the past it was believed that it's good to adjust the instruction method to the learning style (for example, to give more verbal instruction to a people who prefers verbal learning and to give instruction with more graphs and visual material for people who prefer to learn via visual images).  Contrary to the expectations of virtually all, the profile of specific learning styles generally does not account for much of the variation in outcomes. Indeed, interactions between learning styles (such verbalizer versus visualizer) and instructional methods (such as an emphasis on visual versus verbal media) are usually small. Instead, it is advisable to adjust instruction style to the learner's cognitive abilities.  For example, a child with poor reasoning ability can be taught with more examples, and with more explicit explanations.  A child with poor comprehension knowledge can be given texts that are linguistically simpler.  Vocabulary from the text can be taught in advance.  The general information required to understand the text can be also pre-taught.

Reasoning ability and working memory.

Theories that attempt to explain reasoning processes assume there are phases in the inference process that include: A.  Representing the premises in working memory.  B.  Creating mental models of possible solutions that are derived from the premises, holding these models simultaneously in working memory and comparing them.  These processes demand significant working memory resources.  Thus limitations in working memory cause individual differences in reasoning abilities.

Since working memory limits the number of mental models that can be held simultaneously in working memory, people with a limited working memory capacity may not succeed in creating enough models to evaluate the validity of a conclusion. 

Researchers find a large overlap between fluid ability and working memory.  The authors of this chapter argue that this overlap is caused by fluid ability being measured too narrowly and working memory being measured too broadly. 

The authors claim that matrices tests that require nonverbal reasoning predict learning and academic achievement in the real world less well than measures of verbal and quantitative reasoning (for example, verbal analogies tests or mathematical series tests).  Thus, when we assess reasoning with one test like the Raven, or similar tests, we don't represent all aspects of the reasoning process.

On the other hand, working memory tests are more complex than it seems.  These tests involve understanding directions that are fairly complex (for example, in the number – letter series test), forming a strategy and sometimes altering it, performing a difficult task with high demands for attention, and keeping a high degree of continuous effort.  These tasks also require executive functions – monitoring the process, inhibiting wrong responses, and switching sets flexibly.  Thus, working memory tests contain fluid aspects.

To summarize:  reasoning abilities are not static.  They develop through experience and are easier to perform after training.  Individual differences in reasoning correlate significantly with the amount of information people can hold in working memory while performing a manipulation.  The ability to do this depends on the attention resources that a person has, on his degree of familiarity with the information, and on his experience with the performance of the required manipulation.  Thus, prior knowledge and skills determine to a large extent the level of reasoning that can be reached in reasoning tests and in reasoning situations in daily life.

When assessing fluid ability, it's important to include a broad range of tests, not only matrices tests that intentionally decrease the contribution of comprehension knowledge to the reasoning process.  It's important to include tests requiring verbal and quantitative reasoning, which are closer to the way fluid ability works in real life. 

This chapter emphasizes the importance of looking at fluid ability clinically – on the ways it is expressed in tests that do not measure it directly (for example, looking at the child's level of conceptualization when he defines a word, and at his ability to understand implicit meanings in texts).  It also emphasizes the dependence of fluid ability on other abilities like working memory and comprehension knowledge, and the interaction between them. 

The C-test as a measure of comprehension knowledge.

Baghaei, P., & Tabatabaee, M. (2015). The C-Test: An Integrative Measure of Crystallized Intelligence. Journal of Intelligence, 3(2), 46-58.‏

This is a short and clear paper about the C-test, a closure test.  In classical closure tests, every n-th word in the text is omitted.  The child is required to complete the missing words.

The C-test is a new kind of closure test.  The test consists of four to six text paragraphs.  In each paragraph, beginning from the second word in the second sentence, the second part of each second word is omitted.  There are usually 20 to 25 part words in each paragraph.  In order to give the child enough context, the first and last sentence in each paragraph remains intact.  For each correct word that is completed the child gets one point.

Here is an example of a C – test paragraph:

If you were to ask most people who Charles Darwin was, many of them would reply that he was the man who said that we were descended from monkeys. They wo___ be wr___. Darwin d___ no mo___ than sug___ the possi___. What h___ said, a___ proved b___ thousands o___ examples, w___ that ov___ millions o___ years ani___ and pla___ have cha___. This he called evolution.

What does the test measure?

The test measures verbal ability in the first and second language.  Scholars argue that the processing done in this test resembles natural language processing.

The test also measures integration of knowledge about the context (general information), semantic knowledge (knowledge about meaning and content), grammatical knowledge, morphological knowledge (knowledge about the meaning units that make up words), lexical knowledge (vocabulary), and orthographical knowledge (spelling rules).

There's no doubt this test is affected by reading and reading comprehension skills, as well as by fluid ability and short term memory.  The larger the segment of text the child reads in order to complete a word, the higher his score on the C-test.

Difficulties on this test can arise from hasty closure (ending a sentence before it ends), narrow focus (using only the immediate context), wrong spelling, mismatch of singular and plural forms, choosing a wrong word out of a right category, relying too much on genera knowledge (completing according to general knowledge and not according to the text's requirements), inattention to grammatical nuances, automatic completion using highly frequent words and difficulty retrieving lexical items.  It's harder to complete words in long and complex sentences, since it's more difficult to understand these syntactical structures and they load working memory.

Intelligence and Cognitive Abilities Part 7: Crystallized Knowledge

Intelligence and Cognitive Abilities Part 7: Crystallized Knowledge

is here

synonymy

 What’s the difference between a violin and a fiddle? No one minds if you spill beer on a fiddle.

A few days ago I was trying to exit a parking lot but couldn't find the paying machine.  Luckily one of the parking lot employees was just passing by.  He was very helpful: "look over there.  See this white thing with the hole in it? Put your ticket in that  hole."  I looked in the direction he was pointing at, but couldn't see anything white with a hole in it.  At a closer inspection I realized that right in front of me was a white paying machine with a slot for the ticket.  What confused me was that I was looking for a hole, a round thing.

As I'm currently working on a presentation about crystallized knowledge (the 7^th in the nine presentation series about cognitive abilities) this incident made me think about synonymy.

What's really the difference between a slot and a hole? A hole doesn't have to be round.  Is a slot a kind of a hole?  Maybe a slot has to be elongated.  But a hole can also be elongated (to a degree?).  Does a hole have to go all the way through  -  but a slot doesn't?  What about the difference between a slot and a crack? a slot and a slit?

Philip Edmonds and Graeme Hirst address these issues in their paper Near synonymy and lexical choice.  Computational Linguistics 2002 vol.28 no. 2  pp. 105-144

http://ftp.cs.toronto.edu/pub/gh/Edmonds+Hirst-2002.pdf

Here are a few points from this paper:

In a given language, all the words which express neighboring ideas help define one another’s meaning.

Synonyms are words that are identical in “central semantic traits” and differ, if at all, only in “peripheral traits.” But how can we specify just how much similarity of central traits and dissimilarity of peripheral traits is allowed?

On the one hand, any two words are synonyms (because every word denotes a “thing”).   But on the other hand, no two words could ever be known to be synonyms, because, even at a fine grain, apparent synonyms might be further distinguishable by a still more fine-grained representation.

As a possible solution to this problem, the authors introduce the idea of granularity of representation of word meaning.   Granularity is the level of detail used to describe or represent the meanings of a word.    Granularity is essential to the concept of cognitive synonymy, as which pairs of words are cognitive synonyms depends on the granularity with which we represent their meanings.

The meaning of any word must have inherent aspects.   But nuances of meaning can only be fully understood as differences and relations between that word and one or more of its synonyms.  

According to Merriam Webster dictionary, a hole is an opening through something.  A slot is a narrow opening or groove.  A crack is a thin line in the surface of something that is broken but not separated into pieces; a very narrow space or opening between two things or two parts of something. A slit is a long, narrow cut or opening in something.

These four words have the same inherent coarse aspect (opening).  In this level of coarse granularity, they are related words (far synonyms?)   But their fine grained  differences  define each of them precisely.

Why is this important?

Words and concepts are the building blocks of thinking.  Synonymy helps us  make fine word distinctions.  Being able to distinguish between nuances makes us better thinkers.  It also enables us to communicate our meanings with more clarity and precision.  Having a rich "arsenal" of synonyms helps us overcome retrieval lapses as we have alternative forms of expressions.

Another thought:
It's possible that there are individual differences (and possibly developmental differences) in the granularity of the representation of word meanings.  This could explain why some people judge a word pair as synonyms while others don't.  It's possible that the former represent those words with a coarser level of granularity than the latter.