An Overview (by Gove) 

Is important to retain the distinction between primary and secondary deviance, to use Lemert's terms. For labelling theory , primary deviance is less important because it's not the basis of a social process. For them, deviance is not inherent in the act but emerges from interactions between the committer and the responder: it is a matter of audience reaction rather than actor's intention. Hence Becker's famous quote from Outsiders: 

'Social groups create deviance by making rules whose infractions constitute deviance, and by applying these rules to particular people, and labelling them as outsiders... deviance... [is thus]... a consequence of the application by others of rules and sanctions'

(a) the values of the sub culture are not those of the dominant group
(b) there is role conflict [ Anomie?] 
(c) deviance arises from a utilitiarian calculation of gain 
(d) deviants are unaware of the rules [inadequate socialisation?] 

The ability to resist the label has become much more important in understanding patterns of criminality. Public arrest and charging procedures have become a  'right of transition'. Labelling theorists believe that the process is almost irreversible once begun, leading to a self-fulfilling prophecy, and perhaps the development of a subculture to provide an alternative solid identity. 

The main criticisms are: 

(a) The theoretical issues are not very explicit, partly because labelling theory followed symbolic interactionism in undertaking descriptive studies  ('news gathering' about cultural minorities) rather than in tackling problems of evidence, or systematically evaluating labelling approaches against rivals. Further, there was an early commitment to the underdog. Nevertheless, labelling theory is testable in principle -- to see what effects labels have, or how continuous criminal careers actually are. 
(b) Is labelling suppose to be a dependent or independent variable?  [in other words, it is an explanation for crime or does it need explaining itself?]. Sometimes, it seems to be a dependent variable, the results of relations of power, different tolerance levels, social distance between labelers and labelled, the visibility of the act to be labelled, and so on. Status differentials appear to be the most important explanatory variable, and Gove proposes testing this by comparing measures of social marginality with actual deviancy rates -- are the most marginal groups more likely to be labelled as deviant? On the other hand, labelling theory is often seen as an independent variable -- the self-fulfilling prophecy has his own effects. 
(c) The mechanisms of labelling need to be explained in more detail. In particular, a notion of primary deviancy seems to be implicit, if the action of the delinquent is to be separated analytically from official labelling  [as when labelling is in independent variable]. Gove suspects that the deviant activity itself can be an active factor in the process of labelling  [that is, continuing deviancy might attract stronger or more persistent labels, or generally affect labelling behaviour] -- omitting this factor seems odd in an interactionist approach! Similarly, the idea of deviancy as a master status, which is what is supposed to happen after the self-fulfilling prophecy, seems unnecessarily reified. Further : 

1. Types of social reaction are not really explored, but are assumed to be always hostile and exclusionary.
2. Moves back into normality need not be that rare 
3. There is some evidence that deterrence actually works in reforming deviant behaviour 
4. Labelling may have different effects according to whether deviancy is seen as ascribed or achieved 

Labelling theory ignores the issue of evidence. There are some good criticisms of empiricism, but these have led to a bland assertion that criminal and non-criminal action is the same. This ignores the initial conditions that lead to crime. The approach fails to specify the labelling mechanisms, and offers no real research, only anecdotes. It also overpredicts deviancy -- once one is labelled, one will always be a criminal, but everyone has a been negatively labelled at some stage. What of actual evidence ? It is possible that delinquent behaviour itself can be a major factor in official responses, and if this is so, social reaction is of course justified, and official statistics are useful as an indicator of acts that  really are socially harmful. 

Self report research indicates that deviancy appears equally in each social class, but this is not the case for official statistics. This shows that deviancy is extremely common, but does it support the view that the only difference between delinquents and normal people is that the former have been caught -- surely not. Self report data also confirms some of the more traditional correlational material [connections with things like inadequate families?]. Delinquent acts appear on a continuum -- there really do seem to be saints and sinners, and the sinners and crooks do get caught by the police more frequently: the self-fulfilling prophecy is in doubt therefore. 

The selection of delinquents to be processed is also worth investigating. A major variable seems to be the seriousness and frequency of bad behaviour [but whose definition?]. In this sense, the police react rather than proact. Generally, they seem reluctant to label anyone. Those studies of police practices, which show the importance of the demeanor of the suspect, for example, do not really support labelling theory -- they seem to be important as explaining variable treatment within categories of offence. It is true that the police tend to select people who fill out the delinquent stereotype -- someone who is tough, mean, and disrespectful -- but it is partly the fault of the delinquents themselves [charming! And the innocent?] There is some empirical evidence to support police judgements like this -- toughness is correlated with delinquency  [a pretty circular argument here!] Legal variables are as important in the processing of delinquents, and the effects of race or status can be explained in the same way -- these operate within the legal categories, so to speak, and they are empirically related to delinquency. 

Do policies towards delinquents work? The usual statistics seem to show not -- there seems to be a high rate of recidivism. This seems to indicate that official reactions have no effect. This, together with the observation that the general trends towards delinquency die away with maturation raises doubts about the self-fulfilling prophecy. It is not even sure that delinquents suffer particularly from stigmatisation.

Chapter nine Sagarin and Kelly  'Sexual Deviancy and Labelling Perspectives'

Sexual activity is an area of obvious technical interest to theorists of deviancy, since it is very difficult to decide where rules are being broken. In areas such as homosexuality, paedophilia, transvestism, and trans-sexuality, the participants experience labelling as meaning different things -- as blowing cover or  coming out; as being detected, isolated, punished or treated. These differences arise because different power relations are involved in what looks like a unitary process. Further, when labels are rejected by deviants of this kind, this is more complex too -- people are prepared to accept the label as a technical description but not as meaning that they are bad. We thus have a number of possibilities according to these different reactions and experiences -- in fact eight in all [work it out!]. 

This is because an important dimension for sexual behaviour like this is the extent to which it can be kept secret or private. As a result, it might be possible to examine him now detail how labels stick -- it might depend on the degree of visibility, secrecy, or  political activity  (such as involvement in Gay Pride), as well as just 'power relations'. These factors might affect the generalisation of the label too, that is whether one is able to be a voluntary deviant, able to resist a master identity based on one's deviance. It might be possible to test these factors especially at the amplification, or secondary deviance stage. 

The authors feel that labelling theory does fit male homosexuals, less so for transvestites, still less to transsexuals, and not at all to paedophiles. Clearly, ideological goals are important -- transsexuals want to be labelled  [especially if they are seeking surgery, as in the famous study by Garfinkel on Agnes], and here, amplification would help. On the other hand, paedophiles are usually treated as 'sick', and are not subject to subsequent amplification as deviants. [How this has changed! I am transcribing this in the summer of 2000, when there has been in the UK a sustained moral panic and witch hunt directed against paedophiles. Tabloid newspapers have  'named and shamed' such persons, while vigilantes have taken to the streets to demonstrate outside the houses of suspects. The paedophile has become the all purpose threat to social order]. Why are labelling theorists disinterested in these groups? [Because they do not appear as romantic underdogs? Because to take the side of a paedophile would be to invite vigilante action in your own front garden?]. 

In terms of sexual deviance in particular, there appear to be four common positions to explain the behaviour:

(a) deviants and normals are basically the same in terms of features like promiscuity and so on 
(b) deviants are different, but only because they have been repressed and stigmatised -- much as the behaviour of black deviants gets explained by some Marxist critics [the authors' example -- but see my discussion on mugging in Policing The Crisis] 
(c) deviants are different as a result of their own moral choice  [we might say these days as a result of their own politics of identity] 
(d) there are pathological parts of sexual deviance -- especially as in paedophilia 

My comments
I think this raises some interesting specifics about sexual deviants and labelling, but I find this last bit almost impossible to take seriously -- I would have thought that we would need to check the evidence very carefully indeed, to examine matters like definitions, the selection of samples, how factors like ineffective family patterns are operationalized and measured, and so on.