Disclaimers

I am not an expert on cognitive science, psychology, second language acquisition, or anything else this book is about. As a result I may have unintentionally misinterpreted some parts of the book and I am not qualified to comment on the accuracy of any claims made. I will however relate claims made in this book to my own personal experience, and speculate on how the information can be used. Please do your own research before using any information in this review.

This book uses some technical language. I have done my best to check that the terms used in Japanese match the ones I am using in English, but there may be some cases in which I mismatched terms. When I have translated an important word, or a direct quote from the book I will include the original Japanese alongside it.

If you notice any mistakes I made, please let me know and I will correct them.

This review mostly contains information in the book that I personally found interesting. As a result there is a lot of content in the book that is not mentioned in this review.

Introduction

Yasuhiro Shirai is a professor of cognitive science at Case Western Reserve university who has done substantial research into second language acquisition. In his book 外国語学習の科学(EN: The science of foreign language acquisition) he summarized the current¹ theory on how learning a second language works. As someone who studied Japanese for a long time I thought this was very interesting.

The book is less focused on giving specific instructions on a study method, but instead gives an overview of the existing knowledge.

The role of your native language

The book begins with a shocking revelation: it is easier to learn a language that is similar to your native language than one that is very different. For example, for a native English speaker, it is easier to learn a language like German or French than Chinese or Japanese. This might seem obvious, but if you look into why this happens it leads to some good insights.

One of the reasons learning languages that are different from your native language is hard is that when you don't know something, you assume it is like your native language. This is called "language transfer" (言語転移). With similar languages this works in your favor a lot of the time. In a lot of ways German and English are kind of similar² so if you guess it may be right a lot of the time. However, with languages like Japanese and English it will probably work against you. For example, in English when talking to someone else you will almost always say "you" when talking to someone, but in Japanese it would be unnatural to say "あなた" all the time.

Typicality

The book mentions a few ways that this can occur. One section I found particularly interesting was when he was talking about "Typicality" (典型性). A lot of words can be used in a lot of different ways, and these don't have to be the same between every language. The book gives a few examples of this but one that stuck with me was the following three sentences:

1. She broke a cup
2. She broke the record
3. She broke the tradition

When presented with these three sentences English learners were able to identify that the first one was correct, but had trouble with the other two. The first one is the most basic and direct meaning of break. The other two use a more metaphorical abstract meaning. This can be hard to figure out with words in a foreign language, and just because a word in another language translates to a word in your language doesn't mean all the usage is the same.

It is often recommended to use a dictionary which has explanations of the words in Japanese as opposed to one that just provides a rough translation when studying Japanese to help limit the effect of this kind of thing. If you think 置く="put" then you will end up using it in unnatural ways. However, if you look up a word in Japanese then you will get an abstract definition and avoid the artificial precision given to you by the translation.

Early Output

Another topic briefly mentioned was outputting can form habits. He writes the following

He gives the example where Japanese people will say "How do you think" when they are trying to say "what do you think?" which is a direct translation of "どう思いますか". Early output is a controversial topic in the Japanese learning community, and he talks much more about output in future sections, and it will be discussed there. But, I don't want you to read this and think "well that is decisive evidence early output is bad" because the question is more nuanced, and I would prefer you defer judgement for now.

Similar to this, it was mentioned that some study methods lead to more language transfer than others. In particular study methods that focus on translation cause more language transfer, and those which focus on communication cause less transfer.

Sociolinguistic Transfer

One thing that I haven't heard much about before reading this book is Sociolinguistic Transfer (社会言語学的転移). The idea is that to be polite you use certain phrases at certain times. However, the kind of phrases you use will be different in different languages. For example, in English after receiving a compliment you should usually say "thank you", however in Japanese you shouldn't just accept it. So if your language seems otherwise correct you might seem rude if you say "thank you" in Japanese.

Shirai recommends you pay close attention to this because people might be less tolerant of these kinds of mistakes.

Kids vs Adults

One question that often comes up in the language learning community is what is the difference between kids and adults. There are even some people who wonder if it is poosible to learn a language as an adult period. Well the book gives a resounding "we are not totally sure".

This stuff is really hard to study

Imagine you wanted to design an experiment to see if there is a critical period for language acquisition. Well for kids it takes a few years before they are able to speak fluently, so as a fair experiment you need to run it for a few years. Also, some people might be different in other ways: maybe one person has a better study method than another. Or maybe they just spend more time looking up grammar and stuff. So you need to control for that. I think you get the idea: you can't do an experiment.

So what you do is you look for people who are learning a language and try to get evidence from them.

Kids are better

One thing that is agreed upon though is that kids are better at learning languages. However, that is where the agreement seems to end. Some people think that there is a critical period (臨界期) for language actuation, after which you can't get to native level. Others believe you can get to the same final level, but it takes more effort. Amongst those who believe in the critical period there is disagreement about when the critical period is. Also, there is disagreement about how it works with different aspects of language. Maybe there is a critical period for pronunciation, but there is no critical period for comprehension? And assuming there is a critical period, how big is the difference in the maximum level that can be obtained? We don't have decisive answers, but there have been a lot of studies on the subject.

The behavior of second language learners

Most experiments trying to test this have looked at what age people came to the United States, and their final level of English ability as an adult. The idea being, if there is a critical period at some point, the final level will decrease. There have been many experiments showing that older immigrants tend to end up worse at English.

However, one problem with these studies is that the behavior of the immigrants also changes with age. If you were suddenly forced to move to a foreign country, you would probably still watch some TV shows or read some books in English. However, little kids will just watch whatever is on TV. One thing that supports this is there has been a disparity in results when measuring the critical period between immigrants from Asia and immigrants from Russia. Immigrants from Russia seemed to be able to get better at older ages compared to immigrants from Asia. Russian immigrants are mostly white so they might be able to assimilate better and hang out with American people. While Asian immigrants might be more likely to hang out with other Asian immigrants³. This suggests that maybe the brains of adults are not different, but the behavior is different.

Phonetic awareness tests in Taiwan

In Taiwan they did a test measuring phonetic awareness and the age someone started studying English. They found that people who started studying earlier had greater phonetic awareness suggesting that starting younger does have an effect on the final outcome. While people who started studying younger likely studied more, this is a small confounder compared to the potential confounding effects of experiments done in America such as those in the previous section. This suggests, at the very least, that for phonetic awareness age is important.

We only need one counterexample

Observant readers may have noticed that one weakness of the critical period theory is that if we find even one example of someone who got to native level as an adult then there can't be a critical period. This was noticed by researches as well so they spent some time looking for people who got very good at a second language. They found people who got very good, like good enough that they had native speakers around them claim they were native level, but upon closer examination they found small problems in their language ability.

In my opinion this is the strongest piece of evidence for the critical period. In my experience I have met people who have gotten extraordinarily good at English as an adult (including people whos native language was very different like Chinese or Japanese), but no one who is literally perfect. I still think that the gap between children and adults is overestimated by a lot of people, but I think the reality is that there is a difference.

Language learning methods

The input hypothesis and skill building

Shirai presents two philosophies for language acquisition. One is the input hypothesis (インプット仮説), and the other is the skill building theory (自動化モデル).

The input hypothesis suggests that people learn language by hearing comprehensible input. So when you hear and understand something you acquire it. It suggests that language learning is subconscious, and that conscious study of language is not helpful. As evidence there have been people who have not talked at all but were able to speak after getting a lot of input. There have also been study methods built off the input hypothesis that have been effective.

One problem with the input hypothesis is that there have been cases when someone has gotten a lot of input and isn't able to talk in their target language. This is somewhat common amongst children of immigrants who sometimes fail to develop the ability to talk. Also, children who go to school with instruction in a foreign language, but are not forced to speak the second language will be able to understand the second language, but not be able to speak it fluently.

In the skill building theory, knowledge starts off as conscious but after practicing it a lot, it becomes subconscious. In this you learn grammar rules explicitly then after enough practice they become subconscious. This is how we learn most things, for example this is how we learn to drive a car.

One of the major problems with the skill building theory is that language is too complicated to be understood with just rules. There is grammar that even linguists debate the meaning of⁴. So to explain all of the grammar is impossible. However, both native speakers and second language learners are able to pick up a lot of these rules subconsciously. For example, "Open me a beer" is correct, but "Open me the door" is incorrect, however the grammar rule for this isn't usually explicitly thought. So to get this, it is acquired subconsciously.

There are also parts of language that are not governed by rules at all. For most situations there is a correct way of phrasing what you want to say even when there are multiple grammatically correct ways of phrasing it. For example, "Your marrying me is desired by me" is incorrect, but "I want you to marry me" is correct even though both are grammatically correct. Also as discussed in the section on sociolinguistic transfer you need to know very specific phrases to be polite in various situations. In order to understand this you need to get a lot of input.

Most second language acquisition researchers believe something in between the two. In other words getting a lot of input is a necessary condition for language learning, and you should explicitly learn grammar/vocab and practice.

The role of output

He mentions there is disagreement about the effects of output amongst researches. While there is a consensus that input is necessary, there are more varied opinions on output. However, he gives some useful thoughts. One thing he mentions is that it is important to feel that you will need to output. When you feel like you will need to output you will start doing "rehearsals" (リハーサル) in your head where you internally narrate what you might want to say. In scans of peoples brains when people actually speak out loud, and when they talk in their head internally it is fairly similar. So by feeling the need to output, you will start naturally practicing outputting by yourself.

This is not the only benefit. When you are getting input you can usually understand a sentence without understanding all the grammar. For example, if you know "John" is a name, the words "kick" and "ball" then you can probably guess the meaning of "John kicked the ball" without understanding the grammar. The sentence "The ball kicked John" will likely never come up, and if it did, you would probably know from context⁵. However, when you are outputting you can't take a shortcut; you must pay attention to the grammar.

In my experience, when I am speaking Japanese, I will come across something I don't know how to say, and the next time it comes up I think "oh there it is". As was mentioned earlier you can build bad habits, and obviously you can't get new knowledge by outputting. However, it does help solidify the knowledge you do have and is a good thing. The book recommends getting a lot of input and outputting a little bit. Also, it could be good to try to use phrases you know are correct when outputting. Fortunately it is a lot easier to get input than it is to output.

Directly he says this:

There were some other theories on the effect of output discussed in the book, but these were the ones I found most interesting.

So what should I actually do to learn a language?

Note: for this section I also read 英語はもっと科学的に学習しよう by the same author. The information will primarily be from 外国語学習の科学 though.

The last part of the book is dedicated to providing advice on what you should actually do to learn a second language. This part of the book is less focused on explaining the results of specific studies and uses reason to provide recommendations.

Get a lot of input and output a little bit

Comprehensible input is the key to learning a second language, and you will need a lot of it. There are a few strategies that the book recommends for making input comprehensible.

First it recommends getting a lot of input on a specific topic (ideally one you are interested in) until you can understand it pretty well. Because the content is about a topic it will be easier to focus on the language. After you are good there you can take the language knowledge and apply it to other topics. Your knowledge of the subject will also help you understand the language because you can guess what it is trying to say. I remember when I was learning Japanese I watched the Yugioh anime, and read the manga, and by the end of that I could somewhat follow a game of Yugioh in Japanese. Given phrases like "特殊召喚" don't come up that often, but I do think it was helpful.

If you are listening to something it could be effective to first listen to it in your target language then listen to it in your native language then once again in the target language. This will maximize the amount you understand and have a positive effect on acquisition. Also, if there is a text form and an audio form of a piece of media (think audiobooks or a show with subtitles) then you can practice both reading and listening which should boost comprehension. For example, you can listen to something, then read it, then listen to it again. Or you can just read the parts you can't catch through listening.

It is also worth noting that in terms of input something is always better than nothing. So even listening to something passively in the background will have a positive effect.

The key point is you want to get to the point where you are not learning a language but using the language to get information.

You should also try to output at least a little bit every day. This could take many forms; you could keep a diary, record yourself talking, talk with other people, chat over the internet or anything. Also while you are doing this try your best to use correct language. If you have the attitude that so long as you are understood it is fine, you might end up with some bad habits so it is important to try to use correct language.

In particular, you should try your best with pronunciation. As an adult is is pretty much impossible to get perfect pronunciation, but the more you try, the better you will get.

Sentence cards

In the grand scheme of things this is a very minor point, but in the Japanese learning community there is a debate on weather it is better to have a sentence on the front of a flash card or just one vocabulary word. According to the author it seems that sentence cards are better. From 英語はもっと科学的に学習しよう

(Translation included in case you have been learning Japanese using vocab cards 😉)

Conclusion

This book was insanely interesting, and there was a ton of information in it I couldn't fit in this review. The book was well written and engaging, and if you want to pick it up you can check out the official page for it here. Additionally, Shirai has written a lot of other books many of which can be found here.