Last Updated on April 24, 2021

The concept of a reason is ambiguous in the English language. When we talk about the reasons a person has, we are sometimes referring to explanatory or motivating reasons, which are often expressed by saying the “reason that” an agent did or will do something. Other times, we are referring to normative or justifying reasons, which are often expressed by saying the “reason to” do something. Normative reasons are often said to be considerations (expressed as facts or propositions) that “count in favor” of a certain behavior. For example, imagine that John is hungry and is presented with poisonous food which he doesn’t know is poisonous. On the one hand, if he eats the food, we can posit an explanatory reason for why he ate the food; the reason that John ate the food consists in the fact that he was hungry and he believed the food was not poisonous. This would be an answer to the question “Why did John eat the food?” On the other hand, we can also posit a normative or justifying reason to not eat the food; the reason to not eat the food consists in the fact that it is poisonous (combined with the fact that he probably doesn’t want to be poisoned). This would be an answer to the question “Why should John eat or not eat the food?”

The nature of normative reasons is a highly important issue within normative philosophy. Many philosophers have claimed that all normative domains (morality, epistemology, aesthetics, rationality, etc.) are intimately connected to reasons. For example, Jonas Olson (2018) notes that “We talk about normative reasons when we say, for example, that there were no reasons to invade Iraq in 2003; when we say that there is reason to stay sober the night before a fi­nal exam; or when we say that there are reasons to believe that the universe is billions of years old. We can say that the first kind of normative reason is moral, the second pruden­tial, and the third epistemic”.

What is Internalism?

Internalism is a thesis about the necessary conditions for an agent to have a normative reason. It is often formulated as a thesis about practical reasons, which concern reasons for actions or intentions (as opposed to, say, theoretical or epistemic reasons, which concern reasons for beliefs). Following the SEP article on internalism, reasons internalism states that all reasons for an agent to perform an action must bear some special relation to some motivational set of the agent (I borrow the phrasing “motivational set” from the “Internal and External Reasons” chapter in Bernard Williams’ 1979 book Moral Luck). This is a highly generalized formulation of internalism. There are a wide variety of ways to understand internalism, by fleshing out what the “special relation” consists in and by fleshing out what the “motivational set” is taken to comprise (note that internalism as defined here is existence internalism, which is not to be confused with judgment internalism, which is a thesis about the necessary conditions for an agent to sincerely judge or believe that someone has a reason).

The simplest form of internalism states that agents only have reason to do what they desire to do. On this form of internalism, the relevant elements in the “motivational set” would be actions that the agent desires to do, and the “special relation” simply requires that reasons count in favor of performing an action in that set. Almost no one accepts this view, since there can plausibly be reasons for an agent to do things that he does not presently desire. For example, imagine John desires above everything else to eat a delicious burger and free delicious burgers are being offered by a nearby restaurant that John can reach with little cost to himself. It seems like there is some reason for him to go to the restaurant (to grab himself a free burger). However, assume that John is not aware that this restaurant exists, which means John has no desire to grab a free burger from this restaurant. The simplest form of internalism cannot account for the intuition that there is a reason for John to grab a free burger from the restaurant. A slightly improved version of internalism states that agents have reason to do an action that would satisfy a desire that they have. This would be able to account for the above case because John grabbing a free burger from the restaurant would satisfy one of his desires (even though he doesn’t know this). This also seems intuitively plausible for many of the reasons we ordinarily ascribe to people. For example, if Adam desires chocolate ice cream but Bob desires strawberry ice cream, it seems plausible to say that these agents have reason to consume different flavors of ice cream (chocolate for Adam, and strawberry for Bob) precisely because of their particular desires.

More sophisticated forms of internalism appeal to more variations in the relevant “motivational set” and the “special relation”. For example, instead of using an agent’s desires as the relevant elements in an agent’s motivational set, some theories appeal to other kinds of motivational states, such as an agent’s intentions, goals, emotions, or even their overall motivation. Also, the “special relation” is characterized differently as well; e.g. instead of using an agent’s actual motivational features, some theories appeal to the motivational features that an agent would have under certain counterfactual idealized circumstances. For example, some theories posit that an agent’s reasons are determined by the motivational features (desires, intentions, motivations, etc.) that he would hold under full information. This is used to prevent desires from generating reasons if those desires are based on false beliefs. Even more sophisticated theories appeal to the motivational features that an agent would hold while “fully rational”, under reflective equilibrium, under faultless deliberation from one’s current motivations, or countless other possibilities.

There are a variety of different kinds of arguments for and against internalism. Some arguments are extensional, i.e. they begin with considerations that intuitively seem to be reasons and argue that these are (or are not) best explained by internalism. For example, consider the fact that it seems that everyone, independently of their (actual or counterfactual) motivations, has moral reason to take into account the well-being of others. Or consider that it seems everyone, independently of their (actual or counterfactual) motivations, has prudential reason to take into account their future desires. Externalists often argue that these moral reasons or prudential reasons do not depend on an agent’s (actual or counterfactual) motivations, which implies that internalism should be rejected (see section 1.2.2 of the SEP article on internalism for discussion of the tension between moral reasons and internalism; Derek Parfit discussed the tension between prudential reasons and internalism in “Rationality and Reasons“).

There are also non-extensional conceptual arguments for and against internalism. For example, some internalists argue that there is a conceptual connection between normative reasons and explanatory reasons. The motivation for one conceptual argument comes from considering first-personal reasons judgments. Typically, when an agent judges that they have a reason to perform an action, this judgment can be sufficient to motivate or explain their performance of that action. A similar general point is then applied to all actual reasons: if fact p is a reason for an agent A to X, then A’s belief that p must be capable of motivating A to X. But if the Humean Theory of Motivation is true, this is possible only if A has a current desire that would be served by X-ing. In conclusion, an agent has a reason to X only if X-ing engages in some way with an existing desire of the agent (this argument is given by Bernard Williams in “Internal and External Reasons” according to mainstream interpretations. See section 2.1 of the SEP article on internalism for discussion of this argument). I will not evaluate the merits of these extensional or conceptual arguments here.

Analogy with epistemic reasons

I will investigate how well internalism holds as a thesis regarding theoretical or epistemic reasons, i.e. reasons for belief (see here for an explanation of the distinction between theoretical and practical reasons). If internalism seems true regarding epistemic reasons, then that would provide some evidence that internalism is also true regarding practical reasons, as there would be a unified relation that could explain reasons more generally across different normative domains. All else equal, a theory that can explain observations across a wide variety of general contexts in a non-ad-hoc fashion has more plausibility than a theory that must posit many different independent explanations to account for those same observations. On the other hand, if internalism seems false regarding epistemic reasons, then internalists will need to provide an explanation of why actions and beliefs are relevantly different enough to demand different fundamental conditions for their reasons. But it’s not clear to me how such an explanation can be given. After all, the concept of a reason is the same when discussing normative reasons for action and normative reasons for belief. Thus, we would expect the fundamental conditions that govern the presence of reasons to be the same as well. It seems to me then that the viability of internalism rests in part on the success of internalism as a theory regarding epistemic reasons. In what follows, I will argue that internalism is plausible regarding epistemic reasons. Also, I will use “epistemic internalism” to refer to internalism as a theory about reasons for belief, and “practical internalism” internalism as a theory about reasons for action.

Desire-based epistemic reasons

On the surface, it seems immediately implausible to suppose that internalism is true regarding reasons for belief. Whether or not a person has reason for a belief depends on whether that belief fits their evidence. People who form beliefs based on their desires or some other element of their motivational set (rather than evidence) are epistemically irrational. Pragmatic benefits (such as whether something satisfies your desires) might be good reasons for action, but they are the wrong kind of reason for belief (read “The Ubiquity of State-Given Reasons” by Mark Schroeder for a good explanation of the Wrong Kind of Reason problem). There are a few ways that an epistemic internalist might account for this intuitive implausibility. One way is to say that we ascribe reasons for belief to an agent on the assumption that the agent has a desire for truth. Another way might be to say that we ascribe reasons for belief to an agent on the assumption that the agent has some desire(s) that would be served by holding true beliefs. The idea is that there needs to be particular elements in one’s motivational set in order for them to have any particular reason for belief (examples of authors advancing these arguments can be found in section 2.2 of the SEP article on reasons internalism). Thus, epistemic reasons would be internal, but the relevant “motivational set” would be a special subset of an agent’s motivations.

Let us assume that epistemic reasons are internal in the sense just mentioned. This would make epistemic internal reasons function quite differently than practical internal reasons. Practical internal reasons can be generated by almost any ordinary element of someone’s (actual or counterfactual) motivational set. In the case of epistemic reasons, however, ordinary desires will typically fail to generate reason for belief. For example, if a person desires to feel secure, and if they would feel extremely secure if they believed in the existence of a benevolent God, this would not provide them with reason to believe that such a God exists. In other words, if they believed a benevolent God exists because such a belief satisfied one of their desires, then this would intuitively be an example of an irrational belief. This creates a few mysteries for epistemic internalism: why do only certain desires (e.g. desires for truth, or desires that would be satisfied by believing true propositions) have a privileged status as being capable of generating reasons for belief, whereas other desires (e.g. desires for feeling secure) do not have this status? Also, why are reasons for belief generated by such a small subset of specific desires, whereas reasons for action can be generated by a much larger range of an agent’s motivations? This distinction seems unwarranted and ad-hoc (i.e. it is posited merely to patch up unintuitive parts of the theory, even though it is not otherwise motivated by the theory itself).

An epistemic internalist might try to avoid these problems by denying that they assign a special privilege to certain desires, thereby avoiding the charge of being ad-hoc. They might say that, under epistemic internalism, like practical internalism, most ordinary desires do generate reasons. They might say that the desire to feel secure does generate some reason to believe in a benevolent God, but that this reason is outweighed by the reasons generated by other desires (e.g. desires for truth, or desires that would be satisfied by believing the truth). This way, the epistemic internalist avoids the charge of privileging certain desires in an ad-hoc attempt to avoid unintuitive conclusions. However, I believe that there are two problems with this reply by the epistemic internalist:

1. it is not clear why the reason based on, say, the desire for truth should outweigh the reason based on the desire to feel secure in this particular example. On a straightforward account of how to “weigh” reasons generated by desires, the strongest desires are the ones with the most weight. But it seems plausible that (in this particular example, at least) a person might have a stronger desire to feel secure than a desire for truth. Or, more precisely, it seems that the desires that are satisfied when one believes in a benevolent God (e.g. the desire to feel secure) could be stronger than the desires that would be satisfied otherwise (e.g. the desires for truth). Yet, even when this is true, it still seems that belief in a benevolent God (on the basis of a desire to feel secure) is (epistemically) irrational.
2. More importantly, it seems more plausible to say that the desire to feel secure does not provide any reason to believe in a benevolent God. In other words, this desire is not the sort of consideration that is relevant at all to whether someone has reason to believe a proposition. Insofar as someone were to consider their desire to feel secure when reasoning as to whether to believe in a benevolent God, this would be an irrational aspect of their epistemic reasoning, even if this consideration were outweighed by other stronger desires (e.g. desires to form true beliefs).

In general, it seems that the vast majority of desires are fundamentally inappropriate to generate reasons for belief, independent of how strong the desires may be. On the other hand, the majority of ordinary desires do generate some reason for action. This is a problem for a desire-based theory of epistemic reasons because it reveals a sharp disconnect between the conditions regarding reasons for beliefs and the conditions regarding reasons for action. And it doesn’t seem that a desire-based theory of epistemic reasons can explain this disconnect in a non-ad-hoc fashion.

Epistemic analog to the motivational set

Due to the argument given above, I believe that we should conclude that an agent’s epistemic reasons are not based on their desires. Similar arguments can also be used against forms of epistemic internalism that connect epistemic reasons with other ordinary elements of one’s (actual or counterfactual) motivational set, e.g. intentions, goals, emotions, motivations, etc. None of these motivational elements seem appropriate to ground one’s reasons for belief. It may seem that internalism about epistemic reasons is thus implausible, providing evidence against internalism generally.

But perhaps not. The above issues merely indicate that an agent’s reasons for beliefs do not stand in a special relation to the agent’s motivational set. But perhaps all of an agent’s reasons for beliefs do stand in a special relation to a different kind of set, a set which is the epistemic analog to the motivational set discussed thus far; call this the epistemic motivational set. For clarity, from this point on, I will refer to the motivational set discussed thus far as the practical motivational set. The practical motivational set is the same set I’ve discussed so far: it consists of the motivations, desires, goals, intentions, and all of the other considerations which dispose an agent toward certain actions or intentions. Similarly, the epistemic motivational set consists of beliefs, experiences, perceptions, intuitions, and all of the other considerations which dispose an agent toward certain beliefs. Now, a unified internalism for both epistemic and practical reasons can state the following: just as all reasons for action must bear some special relation to an agent’s practical motivational set, all reasons for belief must bear some special relation to an agent’s epistemic motivational set.

This is intuitively plausible to me. Most theories of epistemic justification agree that all reasons for belief must bear some special relation to an agent’s epistemic motivational set. For example, Coherentism posits that one is justified in believing P (i.e. one has reason to believe P) only if P coheres with one’s other beliefs (beliefs are included in one’s epistemic motivational set). On the other hand, Foundationalism posits that one is justified in believing P only if P follows from a foundation of non-inferential knowledge or justified beliefs (non-inferential knowledge includes elements like perception, which falls under the epistemic motivational set). Any considerations which do not bear any relation to an agent’s epistemic motivational set cannot be a reason for that agent to hold any belief. For example, if a certain proposition P is true but it has no relation to an agent’s epistemic motivational set (i.e. their beliefs, perceptions, experiences, etc.), then it seems plausible to say there is no reason for the agent to believe P (despite it being true). For example, it might be true that there are exactly 23,538 planets with life in our galaxy, but (assuming this fact has no influence on our observations or perceptions and we don’t have non-inferential knowledge of this fact) there is no reason for any humans today to believe that; anyone who did believe that would be unjustified (even though their belief would be true).

Global Internalism

We can generalize this kind of reasoning in at least two ways. Firstly, we can apply this kind of framework to attitudes other than belief and action/intention to form a kind of global internalism. For example, an internalist theory about normative attitudes broadly might say: all reasons to adopt attitude X (an intention, action, belief, emotion, desire, fears, etc.) must bear some special relation to the agent’s motivational set for attitude X. The motivational set for attitude X includes all features of the agent that disposes the agent toward or away from adopting the attitude. X in this case can be any attitude for which normative reasons make sense. I believe this class of attitudes is best described by what Thomas Scanlon calls “judgment-sensitive attitudes” in Chapter 1 of What We Owe to Each Other; these are attitudes whose presence in an agent can be influenced by that agent’s judgment regarding one’s normative reasons to adopt those attitudes. Scanlon provides more details on this class of attitudes:

The class of attitudes for which reasons in the sense I have in mind can sensibly be asked for or offered can be characterized, with apparent but I think innocent circularity, as the class of “judgment-sensitive attitudes.” These are attitudes that an ideally rational person would come to have whenever that person judged there to be sufficient reasons for them and that would, in an ideally rational person, “extinguish” when that person judged them not to be supported by reasons of the appropriate kind. Hunger is obviously not a judgment-sensitive attitude; but belief is, and so are fear, anger, admiration, respect, and other evaluative attitudes such as the view that fame is worth seeking

The other opportunity for generalization offered by this global internalism is the ability to adopt a central conception of the “special relation” that can be applied to all judgment-sensitive attitudes. This suggests that there is a common normative faculty articulated by the “special relation” that determines the appropriateness or justification of an attitude, independent of the type of attitude that it is (whether it’s a belief, intention, fear, etc.). This seems to be what we are getting at when we refer to a general faculty of reason, i.e. the faculty under exercise when one is in the act of reasoning. Reason, understood as this general faculty, can be interpreted as a faculty that begins with a starting set of motivations (e.g. desires, goals, etc. for practical reasoning, or beliefs, perceptions, intuitions, etc. for epistemic reasoning) and makes determinations about what final attitudes follow from that starting motivational set. When an agent exercises this faculty, i.e. when an agent is reasoning, the upshot is a judgment about what he has reason to do. The actual reasons that an agent has is determined by what an agent would conclude if he were to exercise this faculty well. In other words, reasons are the upshots of the proper exercise of the faculty of reason. Conceptions of the “special relation” can be seen as attempts to articulate the proper functions of this faculty. Reasons, then, can be said to be the rational extension of one’s starting motivational set (the idea that reasons are intimately connected to rationality is one of the key motivations for internalism).

For example, one might adopt a coherentist theory of rationality. This might be expressed as: a consideration C is a reason for any arbitrary attitude X if and only if C coheres with all of the elements in the motivational set of attitude X. This is a well-known theory of justification when applied to beliefs (i.e. coherentism). But it could also work as a theory of justification when applied to intentions, i.e. one has reason to intend X if it “coheres with” their intentions, goals, desires, etc. (e.g. Michael Smith (1995) has argued that consistency and coherence are standards that justify our desires). Perhaps similar considerations can be applied to aesthetics (which are concerned with reasons to admire or appreciate certain objects) or to our reactive attitudes (e.g. reasons to blame someone, reasons to feel guilt, etc.). Once one fleshes out what coherence (or some other theory of justification) actually consists in, perhaps this single conception of justification can unify reasons across all normative domains.

Degenerate Motivations

I said recently that the motivational set for an attitude includes all elements that dispose the agent toward or away from adopting the attitude. The motivational set will actually need to be more strict in order for global internalism to be plausible. Recall that I argued above that desire-based reasons for belief are implausible. While desires should not constitute normative reasons for belief, they often constitute explanatory (albeit subconscious) reasons for belief. For example, people are highly unlikely to rid themselves of beliefs that are intimately tied to their identity (e.g. political or religious beliefs), despite ample evidence being presented to the contrary. Of course, people tend not to consciously take the fact that a belief is tied to their identity to be a reason to hold the belief, but this fact nevertheless plays a role in explaining their tendency to hold the belief. Thus, the subconscious tendency to protect one’s identity is an extremely strong element in many people’s epistemic motivational set. But this tendency does not provide any epistemic reason for belief. Similar examples can also be found regarding reasons for action. For example, imagine a drug addict hates his addiction but nevertheless constantly injects himself with the drug because of his addiction. This person’s physical addiction would constitute an extremely strong element in his practical motivational set. Nevertheless, unwanted addictions do not provide any practical reason to inject heroin. Generally speaking, it seems that agents can have degenerate motivational features that do not provide them with any normative reasons despite their motivational strength.

One way to account for these degenerate elements is to characterize the “special relation” such that we are only concerned with the relevant motivational features of the agent under idealized conditions. But how are we to characterize the “idealized” conditions? On the one hand, we could characterize the “idealized” conditions in normative terms, i.e. by saying that the idealized conditions are conditions where agents are “fully virtuous” or “fully receptive to their reasons”, etc. The problem with this is that it allows for normative truths that are prior to reasons, but this conflicts with the (what I take to be plausible) idea that all normative concepts can be expressed using the concept of reasons. Further, once we open the door to externalist sources of normativity generally, we also open the door to externalist sources of reasons specifically. On the other hand, we could explicitly exclude all instances of such “degenerate” motivations from the ideal conditions. For example, we could exclude all of what Michael Smith calls “physical addictions”, “emotional disturbances” and “psychological compulsions” from the ideal conditions (see chapter 5 of The Moral Problem). But how do we determine what kinds of motivations are emotional disturbances or psychological compulsions? There doesn’t seem to be a general answer. Further, we would need to exclude more motivations to account for epistemic reasons, e.g. agents in the ideal conditions will not have their beliefs influenced by their identity.

As we plug in more and more constraints into the ideal conditions, this defeats one of the main attractions of internalism, which was that it can provide a simple unified relation to explain the normative reasons we have in a non-ad-hoc fashion. But this explanatory power is lost if the theory has to make endless assumptions about what reasons we have prior to “explaining” them.

Constructivism

So we should not account for the degenerate motivational features of an agent by explicitly excluding specific kinds of motivations from the idealized conditions. Instead, I believe there is a promising strategy for removing degenerate elements that provides a unified account of all normative reasons in a non-ad-hoc fashion. The strategy is to limit the motivational set to an agent’s normative judgments, i.e. the agent’s own personal judgments about the reasons they have. This allows us to accommodate the above examples of degenerate motivations. For example, an agent does not have reason to form beliefs on the basis of protecting their identity because the agent, by their own lights, probably judges that it is irrational to do so (typically, people will deny that their beliefs are based on protecting their identity. Even when a person’s beliefs are based on protecting their identity, they think their beliefs are based on the evidence). Similarly, an agent who hates their drug addiction does not have reason to inject heroin because the agent, by their own lights, does not endorse their drive to inject the heroin, i.e. the agent judges that they don’t have reason to inject the heroin. The motivation for limiting the motivational set to the agent’s normative judgments is based on the insight that not all motivations are rational motivations, i.e. motivations that an agent forms because of what they take themselves to have reason to do. It seems that non-rational motivations – e.g. physical addictions, psychological compulsions, beliefs based on non-evidential considerations, etc. – never actually provide agents with reasons.

Note that in order for this theory to qualify as a form of internalism, normative judgments must be elements of one’s motivational set, which means that normative judgments must be inherently motivating. This is a thesis known as judgment internalism (see here and here for brief discussion), which I personally endorse.

The idea that normative reasons bear a special relation to one’s normative judgments is the basis of a metaethical theory called constructivism. Sharon Street articulates in “Constructivism about Reasons” the central thesis of constructivism as follows: “the fact that X is a reason to Y for agent A is constituted by the fact that the judgment that X is a reason to Y (for A) withstands scrutiny from the standpoint of A’s other judgments about reasons.” Whether a judgment “withstands scrutiny” from other judgments is supposed to be a descriptive fact that can be determined without any normative assumptions. In brief, a judgment withstands scrutiny from other judgments held by an agent if the agent in question could hold all the judgments in full awareness. Note that this allows that agents can be incorrect about their reasons because agents often are not in full awareness of their judgments. Much more needs to be said about this and whether constructivism is a plausible theory. I hope to speak more about the merits of constructivism in a future post. I mention it here only to outline a possible strategy for a global internalist theory of reasons to account for degenerate motivations in a unified, non-ad-hoc fashion.

Concluding thoughts

Practical internalism already has an attraction over externalism in that it can describe a single relation in virtue of which all of our practical reasons are grounded. Externalism, on the other hand, needs to posit multiple independent properties or relations to ground our practical reasons. Further, the fact that internalism can potentially ground reasons in other normative domains (e.g. with regard to reasons for belief) provides very strong evidence in favor of internalism with regard to reasons for action. All else equal, a theory that can explain observations across a wide variety of contexts in a non-ad-hoc fashion has more plausibility than a theory that must posit many different independent explanations to account for those same observations. Thus, if everything I’ve said in this post is correct regarding the viability of epistemic internalism, externalists must show that all else is not equal. That is, an externalist must provide an explanation for why beliefs and actions are sufficiently different such that internalism is viable for the former attitudes but not the latter attitudes. As stated before, such an explanation seems implausible because the concept of a normative reason is the same when we speak of reasons for beliefs and reasons for action. In any event, until such an explanation is provided, it seems that the more plausible route is to adopt internalism as a theory of reasons for action.

Related Works

• Stephen Finlay and Mark Schroeder, “Reasons for Action: Internal vs. External” (2017) – Stanford Encyclopedia of Philosophy entry that defines the terms of the internalism/externalism debate and the main arguments for and against both sides of the debate.
• Bernard Williams, “Internal and External Reasons” (1979) – Surely the most cited argument in favor of internalism. Williams argues that an agent has a reason to X only if X bears there is a “sound deliberative route” to X from his present motivations. The book containing the essay, Moral Luck, can be found here. [archive]
• Christine Korsgaard, “Skepticism About Practical Reasons” (1986) – Korsgaard, a different sort of internalist than Williams, offers some critiques of Williams’ argument. She argues for a kind of Kantian internalism where the capacity of reason alone can motivate action without the presence of any particular desire. [archive]
• John McDowell, “Might There Be External Reasons?” (1995) – Defends the possibility of external reasons against the arguments put forth by Williams.
• Bernard Williams, “Replies” (1995) – Williams responds to the objections put forth by McDowell and Korsgaard.
• Philip Pettit & Michael Smith “External Reasons” – An analysis of the debate between Williams and McDowell. [archive]
• Michael Smith, “Internal Reasons” (1995) – Smith provides an argument for his conception of internalism, which is: an agent has a reason to X in circumstances C if and only if her fully rational self would advice her than fully rational self to X in circumstances C. [archive]
• Stephen Finlay, “The Obscurity of Internal Reasons” (2009) – sketches and critiques the standard interpretation of Williams’ argument in “Internal and External Reasons”. Finlay argues that his interpretation of Williams’ work reveals an argument that not only is more in line with Williams’ intentions but that also avoids many of the flaws of the argument in the standard interpretation. [archive]

Other recommended

• Internal Reasons: Contemporary Readings (MIT Readers in Contemporary Philosophy) – A good collection of contemporary essays from all sides of the reasons internalism/externalism debate.
• Kieran Setiya, “Against Internalism” – Argues that practical irrationality is akin to moral culpability: it is a defective practical thought which one could legitimately have been expected to avoid. It is thus a mistake to draw too tight a connection between failure to be moved by reasons and practical irrationality (as in a certain kind of ‘internalism’): one’s failure may be genuine, but not culpable, and therefore not irrational.
• Julia Markovits, “Why Be An Internalist About Reasons?” – Provides nontraditional arguments in favor of internalism