LOURIE, Circuit Judge.
National Research Council of Canada and its exclusive licensee, North American Vaccine (collectively NRC), appeal from the judgment of the United States District Court for the Southern District of New York holding claims 12 and 25 of U.S. Patent 4,356,170 invalid for indefiniteness under 35 U.S.C. § 112, ¶ 2 (1988), and not infringed by American Cyanamid Co. and Praxis Biologies, Inc. (collectively Cyanamid). North American Vaccine, Inc. v. American Cyanamid Co., 24 USPQ2d 1898, 1992 WL 234591 (S.D.N.Y.1992). We affirm-in-part and reverse-in-part.
BACKGROUND
On February 28, 1991, NRC brought an action against Cyanamid alleging infringement of the 170 patent. That patent, entitled “Immunogenic Polysaccharide-Protein Conjugates,” relates to human infant vaccines against bacterial infections such as meningitis. The principal components of the vaccines are polysaccharide-protein conjugates. The polysaccharide, which is derived from the bacteria against which immunization is sought, has antigenic properties that elicit an immune response. Infant immune systems, however, do not respond well to the polysaccharide unless it is linked to a protein.
While linking of the polysaccharide and protein enhances infants’ immune response to the antigenic polysaccharide, the immune response is hampered by “crosslinking.” The parties indicate that crosslinking obstructs the antigenic sites of the polysacchar-ide, precluding an effective immune response. The amount of crosslinking increases with the number of proteins that are linked to the polysaccharide. The claimed invention purports to avoid the problem associated with crosslinking by linking the protein only to a terminal portion of the polysaccharide, thereby preserving the antigenic properties of the polysaccharide and producing an immune response in infants. The meaning of the expression “a terminal portion” in the claims is a key issue in this ease. The meaning of the term “crosslinking” is also contested.
Trial was limited to the issues of infringement and validity of claims 12 and 25, which are directed, respectively, to a polysacchar-ide-protein conjugate and a human infant vaccine comprising that conjugate. Both are dependent upon claim 11, which reads:
11. An antigenic-polysaccharide-protein conjugate wherein the polysaccharide and protein are covalently linked through a
CH2-NH-protein
linkage to a terminal portion of the poly-saccharide without significant crosslink-ing, said antigenic polysaccharide having a MW [molecular weight] above about 2,000.
(emphasis added). Claims 12 and 25 read as follows:
12. The conjugate of claim 11 wherein the antigenic polysaccharide is selected from the group derived from meningococci, Hae-mophilus influenza, pneumococci, beta-he-molytic streptococci, and E. eoli.
25. A human infant vaccine comprising the conjugate of claim 11 wherein the poly-saccharide comprises at least one of men-[1574]*1574ingococcal polysaccharide and Haemophi-lus influenza polysaccharide.
(emphases added).
A polysaccharide is a sugar molecule having a carbon backbone with hydroxy groups attached, to which proteins may be linked. Linking a protein to a polysaccharide generally involves modifying the hydroxy groups to create one or more active sites and covalently linking the activated polysaccharide to a free amino group of a protein. The result is a polysaccharide-protein conjugate.
Cyanamid’s accused product, “HibTITER,” is a vaccine against Haemophilus influenza (H. flu) type b (Hib) bacteria that can cause meningitis. HibTITER consists of Hib polysaccharides linked to one or more proteins. Specifically, it is comprised of three components referred to as “monomers,” “dimers,” and “trimers.” Monomers, which constitute 50% of the polysaccharide-protein molecules of HibTITER, consist of a single protein linked to a terminal end of one or more polysaccharides; dimers consist of two proteins linked together by polysaccharide; and trimers consist of three proteins linked together by polysaccharides. The protein linkage in each of these components is at a terminal portion of the Hib polysaccharide; none has a protein linkage along the backbone of the polysaccharide. Monomers, having a protein linkage at only one terminal, are referred to as monofunctional. Dimers and trimers, having linkages at both terminals of the polysaccharide, are referred to as difunctional and exhibit what has been called end-to-end crosslinking.
The district court entered judgment for Cyanamid, holding that claims 12 and 25 were not infringed and were invalid for indefiniteness under 35 U.S.C. § 112, second paragraph. NRC appeals from that judgment.
DISCUSSION
A. Infringement
A determination of patent infringement requires a two-step analysis. First, a claim must be interpreted to determine its proper scope and meaning; second, it must be determined whether an accused device is within the scope of the properly interpreted claim. ZMI Corp. v. Cardiac Resuscitator Corp., 844 F.2d 1576, 1578, 6 USPQ2d 1557, 1559 (Fed.Cir.1988). The first step is an issue of law and the second a question of fact. Minnesota Mining & Mfg., Co. v. Johnson & Johnson Orthopaedics, Inc., 976 F.2d 1559, 1570, 24 USPQ2d 1321, 1330 (Fed.Cir.1992) (citations omitted). A determination of literal infringement requires that each limitation in the asserted claims be present in the accused device. Hi-Life Products., Inc. v. American Nat’l Water-Mattress Corp., 842 F.2d 323, 325, 6 USPQ2d 1132, 1133 (Fed.Cir.1988). Only the issue of literal infringement is before us in this case.
Whether HibTITER, which includes monomers, dimers, and trimers, infringes claims 12 and 25, turns on our interpretation of the limitation “linkage to a terminal portion of the polysaccharide without significant crosslinking” in claim 11. It is not disputed that the monomer components in HibTITER meet this limitation. Thus, if, contrary to what the district court decided,1 the monomers themselves meet all the other limitations of the claims, including being a conjugate, infringement of the claims would be established because they contain “linkage to a terminal portion of the polysaccharide without significant crosslinking” and they are quantitatively a significant component of the product. However, at oral argument, counsel for NRC stated that NRC is not contesting this aspect of the court’s decision.2 The [1575]*1575only issue before us, therefore, is whether the other components of the vaccine, the dimers and trimers which have protein linkages at both terminal portions of the Hib polysaccharides and exhibit end-to-end cros-slinking, come within the meaning of the above-quoted language, in which case the totality of the accused product would be infringing.
The district court construed this phrase as being limited to linkage at only one terminal portion of the polysaccharide, excluding all crosslinking, even end-to-end crosslinking. Accordingly, the court held that the claims did not cover the dimers and trimers in HibTITER. NRC argues that the court’s claim interpretation was erroneous and that the limitation requiring “linkage to a terminal portion of the polysaccharide without significant crosslinking” in claim 11 is not limited to linkage at only a single terminal, but includes linkages at either one or both terminals of the polysaccharide. According to NRC, end-to-end crosslinking which involves linkages at both terminal portions of the polysaccharide is not “significant crosslink-ing” for purposes of the asserted claims since it achieves the intended purpose of the invention by avoiding crosslinking along the backbone of the polysaccharide. NRC also argues that the inclusion of references to Hae-mophilus influenza in dependent claim 12 and the specification indicate that claim 11 was intended to be interpreted to include end-to-end crosslinking because polysaccharides from that organism lead to end-to-end crosslinking. Thus, NRC asserts, the properly interpreted claims cover the dimers and trimers in HibTITER. We disagree and conclude that the contested phrase does not include polysaccharide-protein linkages at both terminal portions of the polysaccharide.
A determination as to the scope and meaning of a claim is a question of law which we review de novo. ZMI, 844 F.2d at 1578, 6 USPQ2d at 1559. Therefore, we need not defer to the district court’s claim interpretation. Specialty Composites v. Cabot Corp., 845 F.2d 981, 986, 6 USPQ2d 1601, 1604 (Fed.Cir.1988) (citation omitted). Cyanamid characterizes these issues as questions of fact, imposing on us a “clearly erroneous” standard of review. It argues that the meaning of claim language raises factual questions. We agree that a genuine eviden-tiary conflict underlying the meaning of claim terms raises a question of fact, but the questions before us in this case go to the heart of what the claims mean. Resolution of disputed issues regarding the meaning of contested language is ultimately a determination of what claims mean and what they cover. It is a matter of law for this court to decide without special deference to the district court. Absent “a genuine evidentiary conflict created by the underlying probative evidence pertinent to the claim’s interpretation[,] ... claim interpretation may be resolved as an issue of law by the court taking into account the specification, prosecution history or other evidence.... ” Johnston v. IVAC Corp., 885 F.2d 1574, 1579-80, 12 USPQ2d 1382, 1386 (Fed.Cir.1989) (citations omitted).
In construing claims, we begin with the language of the claims themselves. SmithKline Diagnostics, Inc. v. Helena Lab., Corp., 859 F.2d 878, 882, 8 USPQ2d 1468, 1472 (Fed.Cir.1988) (citations omitted). Claim 11 specifically refers to “a terminal portion”; it does not refer to “any terminal portion” or to “all terminal portions.” While it is generally accepted in patent parlance [1576]*1576that “a” can mean one or more, see Robert C. Faber, Landis on Mechanics of Patent Claim, Drafting 531 (3d ed. 1990) (“In a claim, the indefinite article A or AN connotes ‘one or more.’ ”), there is no indication in the patent specification that the inventors here intended it to have other than its normal singular meaning.
When the meaning of a claim term is in doubt, we look to the specification for guidance. See Hormone Research Foundation, Inc. v. Genentech, Inc., 904 F.2d 1558, 1562, 15 USPQ2d 1039, 1042 (Fed.Cir.1990), cert. dismissed, 499 U.S. 955, 111 S.Ct. 1434, 113 L.Ed.2d 485 (1991). In doing so here, we find no indication in the patent specification that the inventors intended to include end-to-end linkages within the scope of their invention. All references to polysaccharide linkages speak of a linkage, not multiple linkages. For example, the summary of the invention describes the formation of the protein linkage to a terminal portion of the polysaccharide as follows:
We have found it possible to introduce a free aldehyde group into the polysacchar-ide molecule in a terminal location and to specifically couple this aldehyde group to protein without activating other functional groups on the polysaccharide.
Col. 2, lines 33-37 (emphases added.) Similarly, the specification states:
The reductive amination covalently couples the terminal aldehyde group of the poly-saccharide to a free amino group on the protein through a -CHz-NH^protein linkage where the -CHz-derives from the aldehyde group. There is no significant cross-linking by this method....
Col. 3, lines 55-64 (emphases added.) Reference to “the” terminal aldehyde group is consistent with linkage at only one terminal, not at both.
The examples in the specification also support linkage at only one terminal. In teaching the selective periodate oxidation of men-ingococcal polysaccharide, the specification states that “[t]he oxidized polysaccharides now had a terminally-located aldehyde group.” Col. 5, lines 14-15. Use of the term “a” and reference to “group” in the singular shows that the specification teaches linkage at only one terminal portion. Furthermore, in teaching the direct conjugation of the oxidized polysaccharides with proteins, the specification states:
It is unlikely that the reducing terminal sialic acid residue at the opposite end would oxidize to any great extent using these conditions because it exists in solution mainly in its pyranose ring form and such has been found to behave similarly to an interchain residue.
Col. 5, lines 50-54 (emphasis added.) This statement shows that it is “unlikely” to obtain linkage at both terminals. Similarly, with respect to meningococcal group A poly-saccharide, the specification teaches:
[T]he reducing end-group N-acetyl-man-nosamine residue was made into the most susceptible residue by simply reducing it to its open chain N-aeetylmannosaminitol derivative. In this form, the modified group A polysaccharide was selectively oxidized at this residue to generate a terminally-located aldehyde group.
Col. 5, line 67—col. 6, line 5 (emphases added.) Again, reference to “the” reducing end group, “this” residue, and “a” terminal aldehyde are consistent with linkage in the singular.
NRC argues that the “significant cros-slinking” that the inventors intended to avoid was not end-to-end crosslinking, but cros-slinking across the backbone of the polysac-charide molecule. While such an interpretation might make sense as a theoretical concept, it has no support in the specification. The word “backbone” does not appear in the specification, nor does the concept of avoiding linking other than at multiple terminal portions. Contrary to NRC’s argument that “significant crosslinking” is crosslinking along the backbone of the polysaccharide, the specification states:
The reductive amination covalently couples the terminal aldehyde group of the poly-saccharide to a free amino group on the protein.... There is no significant cross-linking by this method ....
Col. 3, lines 55-64 (emphases added.) The above-referenced statement that linkage to the terminal portion results in “no significant [1577]*1577crosslinking” shows that “significant cros-slinking” for purposes of the claims was not intended to mean linking along the backbone as NRC urges, but linking at more than one terminal.
It is the responsibility of patent applicants to disclose their inventions adequately. 35 U.S.C. § 112. There is no such disclosure of the concept of avoiding crosslinking along the backbone in this patent. Thus, an invention of that breadth does not meet the description requirement. See Carmen Industries, Inc. v. Wahl, 724 F.2d 932, 937 n. 5, 220 USPQ 481, 485 n. 5 (Fed.Cir.1983) (“Claims should be so construed, if possible, as to sustain their validity.”) (citations omitted). As explained above, the examples are consistent with monofunctionality, not difunc-tionality. A patent applicant cannot disclose and claim an invention narrowly and then, in the course of an infringement suit, argue effectively that the claims should be construed to cover that which is neither described nor enabled in the patent. One of the inventors, Dr. Jennings, testified that he intended to include difunctional molecules within the scope of his invention. Such after-the-fact testimony is of little weight compared to the clear import of the patent disclosure itself. See Senmed, Inc. v. Richard-Allen Medical Indus. Inc., 888 F.2d 815, 819 n. 8, 12 USPQ2d 1508, 1512 n. 8 (Fed.Cir.1989) (Where meaning of claim term is clear from specification and prosecution history, the inventor’s “self-serving post-hoc opinion testimony on the legal question [ ] whether it should have a different meaning was of little if any significance.”) (citation omitted). •
NRC argues that inclusion of the words Haemophilus influenza in the dependent claims and specification prove that difunctional molecules are within the scope of claim 11 because such organisms lead to difunctional products according to this invention. While it is true that dependent claims can aid in interpreting the scope of claims from which they depend, they are only an aid to interpretation and are not conclusive. The dependent claim tail cannot wag the independent claim dog. Certainly in this case, as we have indicated, there is no further disclosure or description in-the specification of difunc-tional molecules resulting from H. flu or any other organism. Moreover, the district court did not find that H. flu always yields a difunctional result, as NRC asserts, but found that one of the “five or six” H. flu serotypes “is the one which yields the difunc-tional result.” 24 USPQ2d at 1903.
There also was evidence at trial contrary to NRC’s assertion that use of H. flu always results in difunetionality. Dr. Eby, one of Cyanamid’s expert witnesses, testified that the H. flu polysaccharide could be treated to create only a single aldehyde that would result in monofunctionality and that such mo-nofunetional H. flu polysaccharides could have been made by one skilled in the art in 1981 when the patent application was filed. Dr. Eby explained that the reference to H. flu in claim 12 is completely consistent with monofunctionality, thereby providing a basis for its inclusion consistent with the otherwise clear singular meaning of “a terminal portion.” Thus, use of the words Haemophilus Influenza in claims 12 and 25 is not inconsistent with a monofunctional interpretation of claim 11.
Finally, if we were to interpret claim 11 on the basis of NRC’s argument concerning what claim 12 requires, claim 11 would include polyfunctional molecules, which are concededly not within the scope of the invention. That cannot be what NRC intends. We therefore conclude that “linkage to a terminal portion” includes linkage only to “a terminal portion” at one end of the polysac-charide.
Since we have interpreted the expression “a terminal portion” to include only monofunctionality, it is clear that the limitation “without significant crosslinking” excludes from the claims linkage other than monofunctional linkage. End-to-end cros-slinking, consisting of protein linkages at both terminal portions of a polysaccharide, is inconsistent with the clear meaning of the expression “a terminal portion.” Thus, end-to-end crosslinking is “significant crosslink-ing” in terms of this patent and it is excluded from the claims. As noted earlier, the specification does not use the term end-to-end crosslinking or difunctionality, nor does it use any language which conveys that back[1578]*1578bone erosslinking was excluded, but end-to-end erosslinking was not. As for the term “significant,” it is undisputed that 50% of the polysaccharide-protein molecules are other than monofunctional, that they have end-to-end erosslinking. There is no basis for concluding that 50% is not significant. Finally, it is the responsibility of patent applicants to define their inventions adequately. In the absence of any definition of this term in the specification, NRC cannot complain if its after-the-fact definition is not accepted. Thus, the district court correctly construed the claim in concluding that end-to-end crosslink-ing did not meet the limitations of the claims.
Cyanamid, as did the district court, emphasizes an article and speech of Dr. Jennings, one of the named inventors on the patent, in which he described his research on polysac-charide-protein conjugates as directed toward achieving monofunctionality. Jennings made these presentations at about the same time as he filed his patent application. The court concluded that Jennings’ failure to clearly express in his application his intention to claim more than the monofunctional product of his earlier work showed that the claims were limited to the scope of his earlier work. 24 USPQ2d at 1902.
Cyanamid argued before us that the court correctly concluded that the monofunc-tional approach embodied in Jennings earlier work was carried over into his patent. According to Cyanamid, Jennings’ article and speech show the “evolution” of the patent and compel a claim construction to require monofunctionality. While we have affirmed the district court’s interpretation of the claims, we must reject its doing so on the basis of this argument. A patent is to be interpreted by what it states rather than by what the inventor wrote in a scientific publication.
There is no inconsistency between writing a paper (or giving a speech) on a particular embodiment of an invention and then claiming one’s invention more broadly in a patent application. Patents often teach embodiments not carried out in the laboratory; scientific papers rarely do. Thus, it was error, albeit harmless error, for the district court to conclude that the narrow scope of the article limits the scope of the patent claims, which should be judged on their own merits.
Having construed the claims, the next determination is infringement. Accordingly, we must determine whether the court clearly erred in holding that the claims do not read on HibTITER. As stated above, the characteristics of HibTITER are undisputed. The only issue is whether the phrase “linkage to a terminal portion of the polysaccharide without significant erosslinking” reads on the HibTITER product as a whole, which, in view of NRC’s concession at oral argument,3 requires determination whether the phrase reads on a product which includes dimers and trimers as well as monomers. In view of our decision that this phrase covers linkage at only one terminal portion of the polysac-charide and does not include end-to-end eros-slinking, the HibTITER product including dimers and trimers clearly is not within the scope of the claims. Accordingly, the district court did not clearly err in concluding that HibTITER does not infringe claims 12 and 25.4
B. Validity
The district court concluded that claims 12 and 25 were invalid for indefiniteness under 35 U.S.C. § 112, second paragraph, which requires that the “specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter [1579]*1579which the applicant regards as his invention.” Even though we have affirmed the district courts decision of noninfringement, we find it desirable to review this holding of invalidity. See Cardinal Chem. Co. v. Morton Int'l, Inc., — U.S. -, 113 S.Ct. 1967, 124 L.Ed.2d 1, 26 USPQ2d 1721 (1993). NRC is entitled to have the validity of its claims considered rather than have the district court’s judgment merely vacated, leaving a cloud on the patent. Future modification of Cyanamid’s product, or sale by another party, could raise questions of infringement of the claims as we and the district court have construed them. Finally, since we have thoroughly reviewed the case, no added judicial effort is involved.
The underlying basis for the court’s conclusion of invalidity on the ground of indefiniteness was the parties’ stipulation that the five categories of polysaccharides recited in claim 12, and the two categories recited in claim 25, each include a number of members (serotypes) which, when treated according to the teachings of the ’170 patent, would result in polyfunctional molecules, polysaccharides having protein linkages along their backbones. This result is outside the scope of claim 11 and contrary to the teachings of the patent. On the basis of this fact, the court held the claims indefinite, stating that “there are so many different kinds of polysaccharides referred to which would yield different results, polyfunctional, difunctional and mo-nofunctional, some of which results are admittedly contrary to the patent’s teaching.” 24 USPQ2d at 1903. NRC argues that the court’s conclusion was erroneous. Specifically, NRC asserts that Cyanamid failed to meet its burden of proving that the claims are indefinite, and that the parties’ stipulation was an insufficient basis for the court’s holding that the claims are invalid. We agree with NRC that the parties’ stipulation of possible inoperativeness of some species does not constitute an admission that those skilled in the art would not be reasonably apprised of the scope of the claims.
Under 35 U.S.C. § 282 (1988), a patent is presumed valid, and at trial Cyan-amid had the burden of proving facts by clear and convincing evidence showing that the patent is invalid. See Buildex, Inc. v. Kason Indus., Inc., 849 F.2d 1461, 1463, 7 USPQ2d 1325, 1326-27 (Fed.Cir.1988). Compliance with the definiteness requirement is a question of law which we review de novo. Carl Zeiss Stiftung v. Renishaw plc, 945 F.2d 1173, 1181, 20 USPQ2d 1094, 1101 (Fed.Cir.1991). Whether a claim is invalid for indefiniteness depends on whether those skilled in the art would understand the scope of the claim when the claim is read in light of the specification. Orthokinetics, Inc. v. Safety Travel Chairs, Inc., 806 F.2d 1565, 1576,1 USPQ2d 1081, 1088 (Fed.Cir.1986) (citations omitted).
Cyanamid did not meet its burden under the clear and convincing standard to show that one of ordinary skill would not understand what is included within claims 12 and 25. Claim 11, the claim from which claims 12 and 25 depend, requires linkage at a terminal portion of the polysaccharide. The specification is clear that the claimed invention relates to conjugates having a protein linkage at a single terminal portion of the polysaccharide, not at both terminal portions. Thus, while the parties in the midst of a dispute have disagreed concerning the meaning of the claims, the claims are not so lacking in clarity as to be invalid under section 112.
The categories of polysaccharides recited in claims 12 and 25 do include numerous members (serotypes), some of which apparently result in polyfunctional molecules when treated according to the teachings of the patent. The fact that dependent claims include species which might not meet the objects of the invention does not by itself prove that one skilled in the art cannot ascertain the scope of the asserted claims. That objection goes to possible inoperativeness under 35 U.S.C. § 101 or lack of enablement under 35 U.S.C. § 112, first paragraph, neither of which provisions are before us. Moreover, if a species within a dependent claim does not meet the limitations of the independent claim, arguably it is not included "within the scope of that dependent claim.
The law is clear that “‘[i]f the claims, read in the light of the specifica[1580]*1580tion[s], reasonably apprise those skilled in the art both of the utilization and scope of the invention, and if the language is as precise as the subject matter permits, the courts can demand no more.’ ” Shatterproof Glass Corp. v. Libbey-Owens Ford, Co., 758 F.2d 613, 624, 225 USPQ 634, 641 (Fed.Cir.), cert. dismissed, 474 U.S. 976, 106 S.Ct. 340, 88 L.Ed.2d 326 (1985) (quoting Georgia-Pacific Corp. v. United States Plywood Corp., 258 F.2d 124, 136, 118 USPQ 122, 132 (2d Cir.), cert. denied, 358 U.S. 884, 79 S.Ct. 124, 3 L.Ed.2d 112 (1958)). That is the case here.
NRC’s expert, Dr. Schoolnik, testified on cross-examination that, in view of the teaching in the specification to avoid protein linkage along the backbone of the polysacchar-ide, he would know that the polysaccharides recited in claims 12 and 25 refer only to the serotypes that would result in “a clean backbone when the procedures recommended by [the] patent are performed on those structures.” He testified:
[I]t’s understood by everyone working in the field that you first draw out from [the group of bacterial types in claim 12] the particular types that are relevant to infantile meningitis, and then I think, as a scientist, I would look at the structure of the polysaccharides from those types and I would say are they structures that would, when subjected to this process, leave a backbone that’s antigenic and probably, therefore, effective as a vaccine.
We conclude that the claims are sufficiently clear to apprise those skilled in the art of their scope. We thus reject Cyanamid’s argument that indefiniteness “follow[s] directly from stipulated facts,” as that stipulation was insufficient as a matter of law to sustain the court’s holding that claims 12 and 25 are indefinite. Accordingly, we conclude that the court erred in its holding of invalidity.
CONCLUSION
The court did not err in holding claims 12 and 25 of the ’170 patent not infringed, but did err in holding them invalid. Accordingly, the judgment of the district court is affirmed-in-part and reversed-in-part.
COSTS
No costs.
AFFIRMED-IN-PART AND REVERSED-IN-PART.