Age effects on L2 acquisition & learning

Age effects on L2 acquisition
Above: the main square
in Tuscania; the Post
Discussion Topic 6:
The effect of age on L2 acquisition
Opening this discussion carries with it the risk of a never ending
conversation. The issue is complex, and some researchers are very
passionate about it. Nevertheless, in the hope of facilitating new
research, let's give it a shot!
Other research has examined L2 learning in non-immersion settings.
For example, quite a bit of research has examined L2 learning in a
classroom setting. Here the "age" variable might be indexed as the
age of pupils when they first begin to study an L2. The age variable
here is often designated
AOL, age of L2 learning. This designation
is useful because it offers a useful distinction from AOA.
Most L2 age research examines one or more aspects of performance. The aim of most research is to afford a
better understanding of how age affects the
completeness of L2 learning, often designated "ultimate
attainment" or "proficiency". For example, one might ask if groups differing in age also differ in degree of FA
(used to index overall ability to pronounce an L2). Alternatively, one might ask at what age a foreign-accented
pronunciation becomes a likely outcome, even after years of experience using the L2.
It is perilous to completely ignore rate of learning, however, because doing so potentially invalidates the results
of studies that focus on completeness of learning. If we compare groups of Ss who began learning the L2 at the
ages of 10 and 20 years, we might get quite different results after 2 and 20 years of immersion (when the Ss
are 12 vs. 22 years of age compared to 30 vs. 40 years of age). If child and adult learners do differ in rate of
acquisition (assuming the same amount and quality of input) we will need to ensure that the members of both
chlld and adult groups have had the input needed to reach their "ultimate" level of proficiency. How much input?
That's a good question for which, unfortunately, I have no answer. I will simply say that, in my view, if one's aim
is to compare the performance of proficient users of an L2 who differ according to some specific selection
variable (e.g., AOA), it would be better to compare Ss  immersed in the L2 for 20 rather than 2 years.
In 1987 I wrote an article entitled A critical period for learning to pronounce foreign languages? The article
was meant to be provocative and, indeed, it generated a bit of controversy. The key point I tried to make is that
the critical period hypothesis really isn't a hypothesis because there was (in 1987) no adequate way to test it.
One problem I pointed to was the problem of confounds. If L2 learning becomes less successful as, for
example, normal neurological development proceeds, then research that aimed to support the Critical Period
Hypothesis (CPH) would need to compare two groups of Ss differing measurably in some aspect(s) of
neurological development who have been matched for other factors that might influence success in L2 learning
(e.g., overall amount of native speaker input). As far as I know, this has never been done. Perhaps it would be
One colleague in particular seemed to have missed the main point of my 1987 article. He wrote a rebuttal article
in the same journal in which he stated that I had denied the truism that "earlier is better" in L2 learning. Actually,
no. It is obvious to me and everyone else, as far as I know, that earlier is nearly always better than later, at
least after the first few months of immersion. The question that has long interested me is: "Why?"  When
researchers obtain "age effects", which is quite easy to do, their job is then to identify the underlying causal
factor(s) that were responsible for the age effect. This, alas, is quite difficult to do with certainty.
The term "age" in the context of second language (L2) acquisition
research usually refers to the chronological age of people when
they first begin to learn an L2.
In the context of studies examining immersion in an L2 following
immigration, age has been indexed by a variable called
AOA, age of
arrival in a host country where the L2 must be learned for everyday
use and, sometimes, for economic survival. The expectation here is
that the earlier the AOA, the better the performance in the L2.
Still another "age" variable that can be examined in L2 research is
, age at test (i.e., the chronological age of Ss at the time of
testing). This is the age variable examined in gerentology.
We hypothesized that with increasing age, it becomes increasingly difficult for bilinguals to
separate the L1 and L2 subsystems. For Ss' who continued to use the L1 frequently, this
led to the prediction that AAT would affect FA, such that the "relatively old" Ss in our
sample would have stronger FAs than the "relatively less old" Ss. The effect of AAT on FA
proved to be non-significant, however.
Comparatively little research has focused on the rate pr rapidity of L2 learning. Why not? I think it's because
measuring rate of learning necessarily means equating groups for the amount of input received in subsequent
units of time. This is difficult, if not impossible, to do. We can't compare the maximum speed of Car A to Car B if
the two cars travel over different kinds of roads during our test. Alas, the normal road to L2 learning of children
and adults almost always differs in important ways, making direct comparisons difficult at best.
Although the work in my lab focused on highly proficient speakers of an L2 (I would argue that there is no such
thing as "ultimate" attainment), I believe that it is also valuable to study what I call "beginners" and, more
generally, learning in the short term. For example, it would be interesting to compare groups of Ss who differ in
amount or kind of classroom instruction, or groups who have received different kinds of laboratory training (e.g.
Flege 1995 ).
An important limitation of short-term learning studies is that their results cannot be extrapolated to more
experienced L2 learners. Put another way, such studies have little to say about high proficiency L2 learners or
"ultimate" attainment. Concepts drawn from the SLM might be invoked to help motivate hyotheses to be tested
in short-term L2 learning research. However, the results cannot be used to falsify the SLM. (Click
here to see a
conference presentation in which I offered examples of how the SLM might be falsified, assuming that it is
As far as I know, my group is the only one to have examined AAT. In an unpublished
study, we re-tested Italian immigrants who had lived for decades in Ottawa, ON. The Ss
were grouped according to AAT, controlling for differences in AOA. As in previous
research, we recorded the Ss' repetitions of short English sentences following a filled
delay, and then had native English listeners rate the sentences for degree of foreign
accent (FA)  using a 9-point EAI scale.
Write if you have something to say on this topic. Please send a carefully edited text and
permission to publish on this site if you want to make your comments public. Thanks. JEF
Coming up: a brief listening of factors typically confounded with AOA.
A weakness of research carried out within the framework of the SLM is that it provides little information
regarding the time course of L2 speech learning. The SLM proposes that learners will "eventually" establish
phonetic categories for some type "A" sounds, and that when this happens the type "A" sounds will be
new L2 sounds. Research examining short-term learning can help better define which L2 sounds
belong to set "A", that is, which L2 sounds are most likely to become "new sounds". If the research also
compares learners of different ages, the research is likely to provide important new information regarding the
effect of AOL or AOA.
Research examining short-term learning in beginners is well suited to investigate the learnability of various
structures in an L2. An example, a short-term learning study might compare two different kinds of sounds in an
L2. Sound type "A" includes L2 sounds that are perceived to be relatively "distant" in phonetic space from the
closest L1 sound (i.e., perceived to be very dissimilar phonetically from the nearest L1 sound). Sound type "B"