What is the job of a theoretical physicist ? Isn’t physics an experimental science ? Should discoveries not always occur as a result of reproducible, controlled laboratory observations ? Perhaps in no other scientific discipline is the division between “theorists” and “experimentalists” so well-defined and rigid as in physics.
It did not use to be that way. In the days of Isaac Newton, a physicist could (and was expected to) do an experiment and come up with an explanation for its outcome, preferably one (a theory) accounting not just for one particular observation, but for as wide as possible a class thereof. These days, however, as the boundaries of the discipline are constantly being pushed (both toward the very small or to the very big), the sheer complexity of both experiments and theoretical calculations requires a high degree of specialization, which makes it essentially impossible for most mortals to excel at both .
In practice, despite their being called both “physicists”, in principle possessing for the most part the same, broad knowledge of the laws of nature, experimentalists and theorists are typically very different scientists. Physics graduate students who are trained as experimentalists generally take different courses (and even emphasis placed on formal course work is not the same as for their theory colleagues), develop vastly different interests and skills, and I think it is fair to say a rather distinct way of thinking and investigative approach with respect to their theory counterpart.
Not surprisingly, such a pronounced difference in mentality can and does lead to awkward interactions between theorists and experimentalists collaborating on the same problem, as well as, in my opinion, fundamentally different, often conflicting views of the discipline as a whole — but that is for another post, maybe after I retire.
OK, so, what experimentalists do is clear, right ? They set up a laboratory and conduct research by performing measurements of cogent physical quantities on systems of interest — samples of condensed matter, molecular, atomic or subatomic particles, biological membranes, you name it. In some (rare) instances, the experiment can be carried out by a few individuals, perhaps as few as two (e.g., a graduate student and his/her supervising faculty); normally, however, a collaboration among several scientists, often based at different institutions, each bringing his/her own unique expertise will be required.
But what about theorists ? What is their contribution to the whole enterprise (now, now, no foul language, my experimentalist friends…) ? Well, based on my experience I tend to think that we all fall somewhere in an interval where, at one end, you have those who work in close contact with experimentalists. These theoretical physicists (I often refer to these as “Experimentalists who cannot use a screwdriver”) are familiar with experimental methods and measurement techniques, and can make sense of and carry out analysis of raw data coming out of an instrument. More generally, their main contribution in a laboratory setting consists of the identification of physical mechanisms underlying the observed behaviour , consequently providing an interpretation of the data by means of simple (usually phenomenological, i.e., not justified based on first principles, microscopic pictures) models. The task of these scientists is to attempt to explain the outcome of a measurement.
At the opposite end, there are those theoretical physicists (also known as “Failed mathematicians”) who are rather removed from the “trenches”, namely the laboratory. They read articles describing experimental work but do not talk to their experimental colleagues on a daily basis. They are not so interested in interpreting one specific experiment, or making sense of one set of data, as much as in achieving a unifying description, capable of accounting for all observations made on a physical system of interest (or, a class thereof). Their aim is that of extracting, from a relatively large set of observations, the one or few common traits that underlie the same, interesting common behaviour of systems whose detailed compositions may be quite different. They shy away from mathematical descriptions of reality too rich in what they regard as superfluous details, preferring instead to work with simple, toy models — something that often frustrates their experimental colleagues, who do not see the point of trying to understand something that may have little or nothing to do with what they perform their measurements on.
Ultimately, the ambition of these theorists lies not so much in explaining what has been observed, but rather in predicting what has not yet been observed.
Most of us fall somewhere in-between the above two “extremes”, and I can see the point of being a bit of both. On the one hand, falling too close to being a Failed mathematician does entail the risk of losing touch with reality, of becoming enamoured with modelling for modelling’s sake, of drifting into irrelevance (I was going to write “string theory”, not sure why). Conversely, the other extreme, being just the “resident theorist” at the service of an experimental effort, or making one’s main goal that of incorporating into a phenomenological model every bit of new experimental evidence, seems itself limiting (and not likely to originate much insight, in my opinion).
I have to confess, however, that if I were to pick a camp, I would join the rank of the failed mathematicians. Even though physics is an experimental science, which means that connection with an experiment (at least possible in principle) should always be made, theoretical physics should not always be at the service of experiment. I believe that an important part of the job of a theoretical physicist consists of imagining things that a bit off the track, thinking of new phenomena to discover, suggesting new experiments, not just interpreting the results of those already carried out.
But in order to conceive such novel phenomena, sometimes it is necessary to forget about the known ones. There is something to be said for the freedom of thinking about a physical system in abstract, without any connection to previous or ongoing experiments, exploring scenarios that may be unphysical (at the present time, or that we know of, anyway). So, yes, even string theory has a place.
 The only name that comes to mind, when thinking of a prominent physicist who lived in modern times and displayed uncanny ability both as a theorist and as an experimentalist, is that of Enrico Fermi.
 Hey, wait a minute, isn’t that what experimentalists themselves should be able to do ? Just kidding…