Natural language can be used with amazing precision - but most of us don't, at least not most of the time. Precise thinking, however, is absolutely required when learning to program. Python vs. Java may be more about the fluidity of the experience and the burden towards minimal viable programs. Those aspects do affect the learning curve.

When it comes to the use of precision when applying natural language, we need to tread a little carefully. There are two cases that I've seen: folks that use natural language sloppily because their use of language is a reflection of their style of thinking - imprecise and logically disconnected. Learning to program in such cases is somewhere between a possible cure and impossible. But there's also a significant part of the population that is hampered by working in a language acquired relatively late and far from their mother tongue or affected by medical conditions, neither of which necessarily reflects on the sharpness of their thought processes. So, jumping to conclusions in this area is highly problematic. (I am not at all implying that this is something implied by the original question!)

To conclude, I think we need to understand what exactly it is that accelerates learning of programming and how natural language skills (or the lack thereof) relate.

A nice question...

As long as requirements for new systems are collected in spoken language, the imprecise nature will always exist. Will it be worse in English than in other spoken languages? Is it now?

If you want precision, you need to use mathematics - as it is the *most* precise way of describing something. Programming languages are extrapolations of the same mathematics, FORTRAN (FORmula TRANslation) is just the first example.

Natural language can be used with amazing precision - but most of us don't, at least not most of the time. Precise thinking, however, is absolutely required when learning to program. Python vs. Java may be more about the fluidity of the experience and the burden towards minimal viable programs. Those aspects do affect the learning curve.

When it comes to the use of precision when applying natural language, we need to tread a little carefully. There are two cases that I've seen: folks that use natural language sloppily because their use of language is a reflection of their style of thinking - imprecise and logically disconnected. Learning to program in such cases is somewhere between a possible cure and impossible. But there's also a significant part of the population that is hampered by working in a language acquired relatively late and far from their mother tongue or affected by medical conditions, neither of which necessarily reflects on the sharpness of their thought processes. So, jumping to conclusions in this area is highly problematic. (I am not at all implying that this is something implied by the original question!)

To conclude, I think we need to understand what exactly it is that accelerates learning of programming and how natural language skills (or the lack thereof) relate.

This has not been my experience.

A much better measure has been the number of mathematically rigorous

proof based courses: Geometry, Logic and Set Theory, Abstract Algebra. ...

The precision needed in a proof mirrors that needed for a good program.

That is why, before Computer Science departments existed, companies

went to the math dept. for their programmers, and before Computer

Engineering, the Elec. Eng. dept. for their hardware engineers.

Much of the theoretical courses in CS were originally pure math topics.

Automata Theory is semi-group theory, graph theory and boolean algebra

came from math originally, and proofs of the computational complexity

for recursive algorithms invariably lead to math induction.

To settle this question, you would need to perform a study and find a correlation,

and then show that this is indeed a causal relationship and not just

correlated to some other factor.

Spoken language relies on context and shared experiences. It often includes idioms and slang which make it difficult for a non-native speaker to comprehend - let alone being precise like a programming language. Even places that claim to speak the same language often have issues when talking to each other. A common expression is "England and the US are separated by a common language."

One way I've stressed (over the years) to make students understand how precise they need to be, is to focus on the algorithm that they try to implement, rather than the language that they will implement in. In a different discussion here at ResearchGate, I mentioned that I often will assign the "making of a peanut butter and jelly sandwhich" as a homework assignment. I will then come into class with my props (a jar of peanut butter, a jar of jelly, a loaf of bread, and a knife) and attempt to follow the instructions written by the students *exactly* as written. 90%+ of the time, I do not get the desired sandwhich. I have found that this visual aid helps the students to understand what the computer will do, and how they *must* list out every step needed to accomplish the task (being precise), not skipping anything and assuming it will just "happen".

Probably it does affect negatively. However, what is important is to have clear ideas an an efficient way of communicating them. Deficient grammar may produce ineffectual results. For example, the comma usage in the following phrases: "We will eat, mom" vs. "We will mom". The meaning is different. Funny as it may sound, and even if people would argue that in spoken language the -comma- pause will go unnoticed, it remains a fact that in computer languages do occur even disastrous results, just because a single missing (or superfluous) symbol. Taking the comma again. Suppose we are passing a function with two variables, say f(x,y) (notice the comma) from Main, and in the module or subroutine connected to the main program, we declare the receiving function without the comma as "g(x)". Errors will follow suit.

Other contexts may be more tolerant on the way functions and operations are being expressed, but still it is important to be precise (in a broad sense), or as we would say, it is good that a person "knows what is he/she doing". Suppose a similar example as before, and instead of main, we will talk about two modules, one passing a function with two variables, and another, but this time it is *not* of a fixed length, but the programmer has devised a totally dynamical variable. For example, in Mathematica it is possible to create functions which may receive about everything that matches a general pattern, like in f[var_]:= var+var, where the dash after var let Mathematica know that it can be any symbolic expression.

The point is that one can create more general expressions (not instantiated), and still remain unambiguous, i.e., one can be precise in a broader sense, provide care is taken about the logical connections between ideas, and their way of being expressed through a computer grammar.

This is an interesting question. The answer however would require a definition of "precision" when it comes to language. It's not so difficult to define in terms of programming languages. After all, imprecision means the program won't work. But as Jerry Heyman alluded to, spoken language consists of many imprecise facets. Yet, communication still works.

In thinking about research on this matter, I'm left with the ironic thought that "advanced" language is more precise than lay language. That is to say, the register used by the educated will contain terms more precise than the more standard register. Use of that "educated" register, one can hypothesize, will mean you have a broader understanding and grasp of concepts that may help in programming. That isn't to say lack of an educated register prevents a person from being a good programmer; it is meant to say that it may be difficult to create an experiment to answer your question.

Not sure I can answer the precise questions, but I *do* know that success in a computer science program is most accurately predicted by the grade in senior high school English (not math as many would think).

Python would require the same level of 'precision' as any other programming language. Vague code gives wrong answers.

Thank you all for an interesting conversation and discussion. There are lots of factors which may or may not affect the ability of students to learn computer programming. I use language examples because it's something that makes sense to them on the first day of class. The precision/imprecision of a student's spoken languages is probably less of a factor than how precise they are in its usage. For example, Chinese is much less precise than English, but there are many Chinese speakers who are good at programming (and math as well, where the precision may come from).

One aspect of irony is that technology which enables/encourages imprecise use of language -- the quick feedback you can give/get over email reduces the need for the extreme clarity one would expect in snail-mailed letters, and the twitter-verse of 140 character conversations -- may be affecting the ability of people to learn the foundations of said technology.

What is cause and what is effect?

Natural language evolves to suit the needs of the users. If users think precisely and demand precission, they will shape the language to suit their needs. If people user the language sloppily and imprecisely, that use will define the language, and make the language imprecise.

Precise expression in natural language is a skill which can be trained and learnt even in a language riddled with ambiguities such as English. It is not easy to say how much of the skill is in the thinking and how much is in the expression. The part which is thinking definitely carries over to other languages (incl. programming), and maybe even some of the other part as well.

I cannot believe that it is right to blame the language. It is the general lack of emphasis on precise expression which causes the problem. The language itself does not prevent that.

If we were to blame English, I don't think the simple grammar causes any trouble at all. The vocabulary causes more trouble; for two reasons. The large number of speakers in general, and the number of non-native speakers in particular, causes the meaning of words to diverge. The user base is too large to maintain narrow and consistent interpretations. Secondly, the English are particularly fond of subtle understatements and ambiguities, which is rather opposed to precission. No offence intended; indeed I love that aspect of English culture and humour.

After all, it is quite possible to use the same language precisely/academically, casually/relaxed, or ambiously/humouristacally. But I agree that it is a problem that children in school do not sufficiently learn to tailor the style of language to the objective. That is a problem not restricted to English.

Vasile - your last sentence captures the problem completely. Written (and spoken) language can be precise, but if you need people who understand that the requirements that they want have to be exact.

A requirement like "responds within 5 seconds" is precise (to a point). In a previous job we had such a requirement, and in our first iteration of the project, the response time was less than one second. All good, right? The team lead suggested we incorporate a sleep call to make up the difference between how long it actually took and 5 seconds. Why you ask? Because we knew in the second iteration, we would be adding more features/function which would slow the response down. We were still within the 5 second requirement, but now the expectation had been altered - and we would be told it wasn't meeting the requirement. By adding the sleep, we continued to add function/feature through each iteration and the customer never had a "performance" issue.

Hi Stephen,

this is a very interesting question.

Regarding the Python vs Java question, my institute is now teaching Python to first year student and moving to C/Java later in the curriculum.

Python seems to facilitate the students to be productive earlier and focus more on the logic of programming rather than getting familiar with the slightly more complicated C/java syntax and concepts.

The new curriculum has been running for 2 years, so it is to early to see difference in students ability to learn the art of programming.

Regarding your examples of more precise language versus less prices ones, you mentions the fact of having more verbs or more forms as a sign of precision.

I am Italian mother tongue, so Latin-based language like French that you cited.

I found English extremely precise and economical-efficient in expressing concepts.

It is true that English could have more ambiguity, but usually it is not a problem and it is certainty more fast and compact to express a large class of concepts in English.

It is very expressive and efficient.

In my opinion programming languages have the same features, they have to express concepts in an efficient and compact way.

Regarding the precision required , I honestly think (=means I do not have empirical evidence of any sort yet..), that programming abilities are more correlated to the previous background of maths and logic rather than language grammar.

I am performing a study on the performance of foreign non-English speakers in computer science, using the student population at my School of Computing.

The results show how foreign students with poor English have no differences in performance in programming subject when compared to native speakers.

However, this is a slightly different from your problem (but related). Different since foreign students has another mother-language that could affect their programming ability, and their original mother language could have been the key for their programming abilities.

However, it would be possible to correlate performance of English-speaking students in programming subjects with their score in English language (collected from secondary school records or from the quality/presentation style of their written university reports/essays)

My impression is that a lack in teaching maths, logic or procedural writing is the dominant factor, even if grammar is indeed a form of logic.

You cited the Baltic countries, but generally also the teaching of maths and logic is deeper in east-Europe and in general in continental Europe than in the British Islands, so a difference in performance could be attributed to maths rather than grammar.

My experience is that brilliant developers were usually poor in written skills, but it should be indeed something worth to be investigated further.

I cannot recall meeting a 'brilliant' programmer who had poor communication skills, but know lots of mathematicians who do not. And I'd never hire a good programmer who could not communicate well. Let's be clear about my previous comment - it's not about the skill in the language *English* that matters, but skill in whatever language the student speaks. The statistics are clear enough on that score: math skills not as important as language skills in ultimate CS success. I teach programming to artists - their math skills not the best, but wonderful code and ideas come from that group.

Hi James, can you point to statistics about "English" more important than MAths in Computer Science?

My data - covering 2000 students in my University and other Irish University, clearly define the level of Maths points in the final secondary school test as the best predictor of students performance in Computing.

I'll look for it. It;' s a well known result from days gone by here, as we have high school achievement exams and can do the correlations. Was done many years ago (I guess I'm presuming it is still true).

This is not the paper I was thinking of, but was very easy to find on Google: http://www.academia.edu/4601713/Language_performance_at_high_school_and_success_in_first_year_computer_science

Hi James, I'd be very cautious with generalizing such observations beyond specific region and educational system. For example strong correlation between CS and English as First Language may be attributed to stronger scientific background of English students compared to Non-English students considered in this research.

While native English students have some handicap over Non-English students due to higher availability of textbooks and on-line resources it is mostly neglected by national tradition of textbook translation and national CS community size. E.g. in Russia we have a large CS community and lots of translated and local textbooks as well as good level of English knowledge. So Russia is a notorious home of crazy Russian programmers.

There are different kinds of math as well as different styles of teaching natural languages. Programming in its core is a long term planning. I think it is more important than problem solving. It is related to the ability to write long ineligible texts in natural language. If the language skills are scored mostly by grammar accuracy the scores are expected to have low correlation with success in CS.

The situation is similar with math. Oleg Viro, one of the well known Russian math teachers, thinks that Geometry is more important for language precision development in students than other more computation oriented kinds of math. I generally agree with him. So if more weight is placed on computational aspects of math, i.e. problem solving then math scores will be less correlated with CS.

Communication skills are a completely different matter. I know lots of good programmers who have bad grammar and heavyweight or too sketchy writing style. In most cases it doesn't prevent them to be understood especially after a few prompting questions. In many situations misunderstanding is just due the fact that other people are too stupid to understand from the first time. Also many bright students experience diffidence . And it is is the duty of teacher and some times employer to develop their self-esteem.

http://www.pdmi.ras.ru/~olegviro/teach-st/teach-st-g.html

Short answer according to my experience --- YES.

I have seen programmers, and some computer engineers, who don't have enough confidence that they can type correct spelling. A person who cannot type the product key and copies and pastes it in 5 parts, a programmer who thinks he could type wrong name in command prompt, and thus while compilation through command prompt copies the file name through right click--> rename, they are more prone to make spelling mistakes in class names, function names, variable spellings, and thus have less precision in programming.

On the other side, a person conscious with spellings, conscious with minute differences between words, are found to be more precise in my short experience.

What do we mean by precise? Is it a function of what is in my mind or what is in yours?

I thought that precision was a matter of agreement when it comes to interpreting language. In a computer, interpretation is defined in terms of bit patterns and the processes that they are demonstrably (repeatedly ) undergoing?

My question is much more straight forward than you may think. I'll rephrase it.

What is the nature of a word "having a meaning" for example how do I map the meaning of that word in my head to what may be addressed in you head? And what level of accuracy can I expect? (and please leave probability out of it) . Your word "interpretation" glosses over the huge gulf between empirical observations of reality and their re-modelling in terms of language. I might suggest that for a given set of sentences (these for example) there will be as many interpretations as there are readers. (Perhaps in this case only me! hence 1 ... does that make me precise?) The possibility is that collections of words may always be an approximation to some real observables. It may even be that that is viable because for normal usage language conveys meaning at different resolutions. Eg "its a nice day works for many interactions whereas:

"Oh, if you prefer stochastic models, you need a semantic space with a metric, and what we want for "precise" is that the distribution of probable semantics should have a small standard deviation."

Requires a narrow or high resolution interpretation

The word precise may well be incorrectly applied to language and hence we need it's meaning to be sure. When we add the concepts of generality and resolution to language we might see it in terms of a better model.

The agreement bit is needed because of the mapping of symbols (words) to empirical observations. This being the only way of confirming and hence evaluation precision.

If you disagree with me then I contend that you have different experiences from me and it is to be expected but in my mind its you that is wrong ...but in your mind I expect you to be correct. Now try defining precision.

A linguistic academy has limited power. People do not necessarily care enough or know enough to know what they say. If they do, they may object, and even if they agree, they may deviate from the norm occasionally. Even French see new creations of words created by arbitrary speakers, and Esperanto, entirely created about a century ago, very quickly saw blatant deviations establishing themselves as common use.

In any case you have a Catch 22. If words are ambiguous or vague in their meaning, you may have to define them. To do that, you need to use … words … which are ambiguous or vaguely defined. Are we into Gödel's incompleteness theorem again?

We seem to have the concept of agreement wrt meanings (on which we appears to agree!)

Now all we need is the acknowledgement of sampling/resolution (accuracy)

(So its not solved yet)

All input to our senses is sampled we can't retain a continuous function of time. Sampling is a processes which includes quantisation and this implies a degree of precision. Language processes the classes from this input and for us associates then to form a context (current state which now includes time ) The progression of the context plus internalised feedback (otherwise known as thought) provides us with two capabilities 1) Communications, 2) Predictive modelling (planning) The communications part only works if there is a common basis (agreement) however there is no requirement for the basis to be in accordance with reality. It is the modelling part in association with testing (against predictions) which forces coherence with reality. The two uses benefit if they use the same basis.

What do we mean by precise?

For example the precision of a sentence could be defined in terms of resolution and scope rather than accuracy. For the domain of experience that would be hard to do. But is may well be realiseable for limited domains such as mathematics and computers.

The original question was about students being precise in language and hence computing. It is much easier to be precise in computing than it is in language . The domain extent is greatly restricted. The models are therefore limited. However the range of uses that the systems can made to handle does actually span our real-time world. So to write systems orientated around computers we might need to be precise and versatile.

Example...defining a real-time response to (lets say) a torpedo attack on sonar-based command system ( I have picked on my last role as a contractor to get a real example)

We first find that the domain includes underwater warfare, sonars, countermeasures, environmental conditions etc. This means that precision really does depend on sampling and quantisation (sonar signal processing), and solutions must be approximate and in fact they can evolve as the engagement unfolds. This defines a class of system which one can describe as non-deterministic. We don't like non-deterministic solutions. They are regarded as imprecise, but humans and real-times systems are non-deterministic. Their advantage is in their adaptability. The best programmers I have ever met are first adaptable then logical (precise).

To handle: requirements where precision can be limiting. To handle contingencies, imagination is called for and to match performance to real world events ... often only approximation is fast enough.

A key ingredient of programming is linguistic ability not for its precision but for the ability to model the system under construction. I believe that this is more important than the re-representation in any chosen computer language.

If the students natural language forces them to be precise, and this has a positive effect on their ability to write good programs, then this should be observable when you compare Asian versus Western performance. E.g. Japanese allows to be much less precise than English. It allows and invites to leave much more unsaid and leaves interpretation to be context and even intonation dependent. I call it a language for the powerful (who can select the interpretation later when and how they see fit). It also allows Japanese poetry to achieve levels out-of-reach for western languages.